No, it's not a robot: prompt writing for investigative journalism

What I’ll talk about:
https://www.bcu.ac.uk/media/research/sir-lenny-henry-centre-for-media-
diversity/blog/six-principles-for-responsible-journalistic-use-of-generativ
e-ai-and-diversity-and-inclusion
Why generative AI is full of %@!*
(And how to use it)
Prompt design — and dangers

Open one of these tools:
ChatGPT: chat.openai.com
Google Gemini: gemini.google.com
Claude claude.ai
‘Custom GPTs’: chatgpt.com/gpts (limits)
These are Large Language Models (LLMs)

It can be an issue: e.g. housing
It can be an event: e.g. doctors striking
It can be an angle: e.g. wide variations in
how much school sports kit costs
Pick a story:

“What ChatGPT deals
in is not truth; it is
plausibility.”
Tim Harford

“Bullshitters are worse
than liars. They don’t care
whether something is true
or false. They care only
about rhetorical power — if
a listener or reader is
persuaded.”
Harry Frankfurt, author of On Bullshit, paraphrased

Not all our work is about facts.
But…

Journalism is also about creativity.
But…

Succinctness and clarity.
But…

Coding.
But…

Coding.
And reflexivity.
But…

“They’re language models,
meaning they’re really
good at tasks involving
language. But they’re not
fact models, or verification
models, or math models.”
Gina Chua

The editorial workflow
● Pre-production: loosen up ideas, break down and distil
systems and rules, aid with planning.
● Research: scope (diverse) sources, explore documents, form
advanced searches, and write/fix code for scraping and
analysis
● Production: identify jargon, bias
● Post-production: optimise and reversion

Generate 10 story ideas about
[ENTER YOUR TOPIC HERE]

Sure, here are 10 story ideas centered around the concept of a "climate visa":
1. A New Beginning: Follow the journey of a family from a small Pacific island that has been
granted climate visas to relocate to New Zealand due to rising sea levels. Explore their
struggles with leaving their home, integrating into a new culture, and the emotional impact
of becoming climate refugees.
2. The Climate Visa Activist: A young environmental lawyer fights to establish a new
international policy for climate visas. The story follows her battles in court, the opposition
she faces from powerful industries, and her personal sacrifices, leading to a landmark
global agreement.
3. Climate Visa and the American Dream: A Midwestern American town facing severe
drought opens its doors to climate visa holders from Sub-Saharan Africa. The influx of new
residents brings cultural exchanges, economic challenges, and a renewed sense of
community as they all work together to rebuild.
4. The Last Generation: Set in 2050, a journalist documents the final years of a village in the
Arctic Circle before…
Generate 10 story ideas about [the climate visa]
��

Certainly! Here are 10 feature ideas about the climate visa:
1. "A New Home: Personal Stories of Climate Visa Holders": Dive deep into the lives of
several climate visa holders, exploring their journeys from climate-stricken homelands to
new beginnings in host countries. Highlight their struggles, adaptations, and hopes for the
future.
2. "The Making of a Climate Visa: Behind the Policy": Explore the political and
bureaucratic processes involved in creating and implementing climate visa policies.
Interview policymakers, activists, and legal experts about the challenges and
breakthroughs in establishing this new form of migration.
3. "Cultural Exchange: How Climate Migrants Enrich Their New Communities":
Highlight the positive cultural impacts of climate visa holders on their new communities,
showcasing stories of cultural exchange, integration, and mutual enrichment.
4. "Economic Boon or Burden? The Financial Impact of Climate Visas": Analyze the
economic effects of climate migrants on local economies, examining both the challenges
and benefits. Include perspectives from economists, business owners, and the migrants
themselves.
OK. Generate 10 ideas for features about [the climate visa]
��

Kate Moran: CARE: Structure for Crafting AI Prompts, 2024
Francesco Marconi, AI Prompt Writing for Journalists, 2023
Prompt design is a CAREful skill.
● Context: describe the situation (role, employer, events)
● Ask: request specific action
● Rules: provide constraints (audience, story type, length)
● Examples: demonstrate what you want
+ Iterate, refine, and combine outputs (break down into
steps) The last 10% is 90% of the effort

To improve this prompt, you could consider the following suggestions:
1. Be more specific about the context of the "climate visa".
2. Define the scope or purpose of the features you're looking for.
3. Specify any constraints or requirements for the ideas.
4. Indicate the target audience or beneficiaries of these features.
5. Request a brief explanation for each idea.
Here's a revised version of the prompt:
"Generate 10 innovative features for a hypothetical 'climate visa' program designed to address
global climate change. For each feature, provide a brief explanation of its potential impact.
Consider aspects such as eligibility criteria, benefits, duration, and obligations for visa holders.
Ensure the ideas are diverse, covering environmental, economic, and social dimensions of
climate action."
How can I improve this prompt:
Generate 10 ideas for features about the climate visa

https://www.youtube.com/watch?v=afDhG8swkuM&t=700s

Investigating the UK school system for an audience aged 20-40, who are likely to include young parents, recent
graduates, and individuals interested in societal issues, would benefit from focusing on current and pressing issues
within education. Here are several suitable topics for investigation:
1. Impact of COVID-19 on Education: Investigate how the pandemic has affected schools, students, and
educational outcomes. This could include issues like remote learning effectiveness, disparities in access to
technology, and mental health implications for students and teachers.
2. Student Mental Health: Explore the mental health challenges faced by students, including stress, anxiety, and
access to mental health support within schools. Investigate whether schools are adequately equipped to address
these issues.
3. Diversity and Inclusion: Investigate how schools are promoting diversity and inclusion among students and
staff. This could include examining policies on racial diversity, LGBTQ+ inclusivity, and accommodations for
students with disabilities.
4. Student Debt and Financial Literacy: Investigate the impact of tuition fees, student loans, and financial literacy
education within schools. Explore how well-prepared students are for managing finances post-education.
5. Employability Skills: Explore how well schools are preparing students for the job market. Investigate initiatives
related to vocational education, work placements, and partnerships with businesses.
6. Digital Skills and Technology Integration: Investigate how technology is used in classrooms and whether
schools are adequately preparing students with digital skills. This could include issues like access to technology,
digital literacy, and cybersecurity education.
7. Environmental Education and Sustainability: Investigate how schools are addressing environmental issues
and promoting sustainability. This could include curriculum content, eco-friendly practices within schools, and
You are an [investigative] journalist working for a [national broadcaster] aimed at an audience
[aged 20-40]. Identify parts of the [UK school system] that might be suitable for an investigation.

Paul Couvert /
Tip: make it visual with Mermaid
1. Prompt: “Write code in Mermaid to explain ____”
2. (Make it human-centred, e.g. “How a person moves
through the system”)
3. Paste the code generated on mermaid.live/edit
Or: use the Custom GPT Diagrams: Show Me

Write code in Mermaid to show how a
person moves through the [UK criminal
justice] system from the point at which
[a crime is reported] to the point of
[sentencing]
*Starting
point!

In the UK, rules regarding school uniforms are typically set by individual schools, but they must adhere to general
guidelines and legal considerations. Here are some key aspects and rules related to school uniforms in the UK:
1. Autonomy of Schools: Schools in the UK have the autonomy to set their own uniform policy, including items
such as colors, types of clothing, and any additional requirements like blazers, ties, or specific shoes.
2. Legal Requirements: While there is no national law mandating school uniforms in the UK, schools have the
authority to enforce their own uniform policies as long as these policies are reasonable and comply with relevant
laws.
3. Equality Act 2010: Schools must ensure that their uniform policies do not discriminate against any particular
group protected under the Equality Act 2010, such as race, religion, disability, or gender. For example, policies
should accommodate religious dress or allow flexibility for disabled students.
4. Affordability: The Department for Education's guidance suggests that schools should consider the cost
implications of their uniform policies on parents and ensure that uniforms are affordable and value for money.
5. Consultation: Schools are encouraged to consult with parents, students, and staff when setting or changing
uniform policies to ensure they reflect the views and needs of the school community.
6. Exceptions and Special Circumstances: Schools may make exceptions or accommodations for students in
certain circumstances, such as financial hardship or medical conditions.
7. Enforcement: Schools have the authority to enforce their uniform policies and may impose sanctions, such as
detention, for non-compliance. However, policies should be applied consistently and fairly.
8. Public Opinion and Debate: Uniform policies sometimes become a topic of public debate, especially when they
are perceived as too strict or when they clash with individual freedoms or cultural practices.
You are a [UK journalist] writing for a [rural] audience looking for [feature] story ideas relating to
[education]. What rules do [UK schools] have to follow in relation to [school uniform]?

_Planning_
‘[Telegraph] staff were also permitted to use AI
“to predict story development and lines that
might be pursued on a long-running story”’

1. Impact of COVID-19: Investigate how the pandemic has exacerbated the shortage of NHS dentists and explore
recovery strategies.
2. Workforce Challenges: Analyze the reasons behind the declining numbers of NHS dentists, including
retirement trends, job satisfaction, and recruitment difficulties.
3. Government Initiatives: Review current and proposed government policies aimed at increasing the number of
NHS dentists, such as training incentives, retention strategies, and funding allocations.
4. Technology Integration: Explore how advancements in dental technology and telemedicine could alleviate
workforce pressures and improve patient access to dental care.
5. Regional Disparities: Investigate geographical variations in dentist availability across different regions of the UK
and the impact on local healthcare services.
6. Role of Dental Hygienists and Therapists: Assess the potential role expansion of dental hygienists and
therapists in delivering primary dental care and supporting NHS dentists.
7. Patient Perspectives: Survey patients and community groups to understand how the shortage of NHS dentists
affects access to dental care and health outcomes.
8. Educational Pathways: Examine the capacity and effectiveness of dental education programs in meeting the
demand for NHS dentists, including barriers to entry and student retention.
9. Private Sector Influence: Investigate the interaction between NHS dental services and the private sector,
including its impact on workforce retention and patient care.
10. Future Projections: Consult with experts to forecast future trends in dental care demand, workforce supply, and
healthcare policy implications for NHS dentists.
You are a journalist writing for a specialist health magazine aimed at people working in the UK health
sector. The number of NHS dentists has been declining for many years now. Make a list of
potential developments in this issue, or future storylines to plan ahead for. Do not write a story.

1. Research and Preparation:
● Understand the Clinic: Research the private clinic thoroughly, including its history, services offered, patient
demographics, and any affiliations with NHS services if applicable.
● Identify Key Contacts: Find out who the appropriate person is to contact for interviews, such as the clinic
manager, head dentist, or PR representative.
2. Craft Your Pitch:
● Clear Objective: Clearly define the purpose of your story and why you are interested in interviewing someone
from their clinic.
● Value Proposition: Highlight how participation will benefit the clinic, such as raising awareness of their services
or presenting their perspective on an important healthcare issue.
3. Professional Communication:
● Initial Contact: Reach out via email or phone with a concise introduction of yourself and your publication,
explaining your interest in speaking with them.
● Respect Their Time: Acknowledge that you understand they are busy and propose a specific time frame for the
interview.
I am an investigative journalist planning to approach a private clinic for an
interview for a story about the growing influence of the private sector on NHS
dentistry. Suggest strategies to consider in making my approach.

Tip: teach your assistant
● Explain the 8 angles approach to data journalism
● Or the Iceberg Model (systems thinking)
● Or the ‘5 Whys’
● Story-based inquiry
● Eisenhower Matrix

You are an editor in a data journalism team at a UK news website. I am
going to give you some information about types of story angles that can
help you suggest story ideas based on any dataset.
Here are the first four angles. Please just confirm you understand this, and then I will give you some data
to generate ideas for.
Scale stories — those which identify a big problem, or the size of an issue which has become topical —
are often the easiest and quickest to do.
Change stories are even more interesting — if you do find that things have changed. The release of
unemployment figures, for example, will see reporters analysing the figures to see if unemployment has
gone up or down. A lack of change can be newsworthy too: if the government has pledged to cut crime
and it’s not changing, then that’s important to report.
Ranking stories are all about what, who, when and where are the best or worst when it comes to a
particular metric. It might also be about how a category or place relevant to the audience ranks against
others.
Variation stories focus on areas where we would expect people to be treated equally, or have equal
access to certain facilities — and reveals that people are not treated equally, or don’t have equal access.
These are often called ‘postcode lottery’ stories

OK, here are the first two rows of a dataset on the gender pay gap at each company in the UK. Use
the four angles I explained to suggest 10 story ideas for this data:
EmployerName Address CompanyNumber SicCodes DiffMeanHourlyPercent
DiffMedianHourlyPercent Women earn more than men? DiffMeanBonusPercent
DiffMedianBonusPercent MaleBonusPercent FemaleBonusPercent MaleLowerQuartile
FemaleLowerQuartile MaleLowerMiddleQuartile FemaleLowerMiddleQuartile
MaleUpperMiddleQuartile FemaleUpperMiddleQuartile MaleTopQuartile FemaleTopQuartile Female
top quartile 50 or above No link? Name Position CompanyLinkToGPGInfo
ResponsiblePerson EmployerSize CurrentName SubmittedAfterTheDeadline DueDate
DateSubmitted
"Bryanston School",Incorporated "Bryanston House,
Blandford,
Dorset,
United Kingdom,
DT11 0PX" 226143 85310 19.6 32.5 FALSE 0 0 32.4 67.6 38.2 61.8
44.9 55.1 53.3 46.7 FALSE FALSE Michael McGovern Chief Operating Officer
https://www.bryanston.co.uk/employment Michael McGovern (Chief Operating Officer) 500 to 999
"Bryanston School",Incorporated FALSE 05/04/2019 00:00:00 03/04/2019 13:43:34
Example

Danger ahead: knowledge cutoff (and RAG)
ChatGPT/Gemini is only trained on data up to a particular date (e.g. Jan 2022,
April 2023, Nov 2023 depending on platform/version)
This is their ‘knowledge cutoff’.
Retrieval Augmented Generation (RAG) is a way of improving responses by
augmenting the query with other information
AI tools like Perplexity and Factiverse’s custom GPT use RAG to augment your
prompt with results of current searches. Others like AnythingLLM (“open-source,
private, chat with your documents”) augment with documents you connect it to.

As of my last update in January 2022, I don't have real-time
information, but I can give you an idea based on recent trends.
French politics have been dynamic, typically involving debates over
economic policies, social reforms, and EU relations. President
Emmanuel Macron has been a central figure, implementing reforms
aimed at liberalizing the economy while facing opposition from
various political factions. Issues such as immigration, climate
change, and national security also play significant roles in shaping
political discourse. For the latest updates, it's best to consult recent
news sources.
What is currently happening in French politics

1. Academic Experts:
○ Contact academics specializing in healthcare policy, economics, and dentistry. They can provide
data-driven insights into trends, challenges, and impacts of private sector involvement in NHS dentistry.
2. Government and Regulatory Bodies:
○ Speak with representatives from NHS England, the Department of Health and Social Care, and relevant
regulatory bodies like the Care Quality Commission (CQC). They can provide official perspectives on
policies, regulations, and oversight of NHS-private sector interactions.
3. Healthcare Professionals:
○ Interview NHS dentists, dental hygienists, and therapists working both in NHS practices and those who
have experience or insights into private dental clinics. Their firsthand experiences can provide a frontline
view of how the private sector impacts patient care and workforce dynamics.
4. Patient Advocacy Groups:
○ Reach out to patient advocacy organizations focused on healthcare access and patient rights. They can
offer insights into how the availability and quality of dental care provided by the private sector affect
diverse patient groups, including marginalized communities.
5. Trade Unions and Professional Associations:
○ Contact dental trade unions (e.g., British Dental Association) and professional associations representing
dentists and dental care professionals. They can provide perspectives on workforce issues, including
recruitment, retention, and the impact of private sector competition on NHS dentistry.
6. Community Representatives:
○ Interview community leaders or representatives from diverse communities within the UK. They can share
Suggest sources for an investigative feature on the
[influence of the private sector on NHS dentistry].

You are a [UK journalist] working on an investigation into [schools' contracts with
suppliers of school uniform and sports kits]. Write an FOI request that can be sent
to [each school] asking for details of the contract - anticipating potential
exemptions that they may try to use to refuse the request
I am writing to request information under the Freedom of Information Act 2000 regarding the contracts your
school has with suppliers of school uniforms and sports kits. Specifically, I am seeking the following details:
1. Names of Suppliers:
○ The names of all companies or organizations currently contracted to supply school uniforms and
sports kits to your school.
2. Contract Terms:
○ The start and end dates of each contract.
○ The total value of each contract.
○ Any financial incentives or rebates received by the school from the suppliers.
○ Details of any exclusivity agreements within the contracts.
3. Procurement Process:
○ A description of the process used to select the current suppliers, including any tendering or
competitive bidding process.
○ Copies of the tender documents or requests for proposals (RFPs) issued by the school.
4. Price Lists:
○ The price lists provided by the suppliers for school uniforms and sports kits.
5. Contractual Obligations: Example

Should you feel that any of the information requested is exempt from disclosure, please
provide a full explanation of which exemptions apply and why, including a consideration
of the public interest test.
In particular, I anticipate that exemptions might be cited under Sections 41 (Information
provided in confidence) and 43 (Commercial interests) of the Act. In such cases, I
request that you:
● Specify the nature of the information withheld and the reasons for applying the
exemption.
● Consider the public interest arguments in favor of transparency and
accountability, particularly given the impact on parents and students.
● Provide as much of the requested information as possible without compromising
the exempted material.
If any part of my request is unclear, I would be grateful if you could contact me as soon
as possible for clarification. I am willing to refine or narrow the scope of my request to
assist in its fulfillment.
Example

1. Government and Official Reports:
○ NHS Digital: Annual reports on NHS dental activity and workforce statistics.
○ Department of Health and Social Care: Policy documents and reports on NHS dentistry.
○ Office for National Statistics (ONS): Data on healthcare expenditure and private sector involvement.
2. Scholarly Articles:
○ PubMed and Semantic Scholar: Research papers on the impact of privatization on dental care quality
and accessibility.
■ Example: "The impact of private sector involvement on NHS dentistry in the UK" (Check citation
counts on Semantic Scholar for credibility).
○ Journal of Dental Research: Articles on healthcare policies affecting dentistry.
3. News and Media:
○ The Guardian, BBC News, and The Telegraph: Investigative articles and news reports on NHS dentistry
and private sector influence.
○ Health Service Journal (HSJ): In-depth articles and analysis on NHS policies and private sector
partnerships.
4. Expert Opinions and Interviews:
○ British Dental Association (BDA): Position papers and statements on the state of NHS dentistry.
○ Interviews with NHS dentists, private dental practitioners, and healthcare policy experts.
5. Case Studies and Patient Stories:
○ Collect testimonials from patients who have experienced both NHS and private dental care.
○ Case studies from regions with significant private sector involvement in dental services.
Suggest sources for an investigative feature on [the
influence of the private sector on NHS dentistry]. Factiverse custom GPT response

“Elephant in the room”
“Elephant in the room
photorealistic style”
“Elephant in the
newsroom photorealistic
style”
Notice anything missing in these images?

“Elephant in a newsroom
photorealistic. People in the image
must reflect the diversity of society”
“Elephant in a newsroom with
diverse people - photorealistic
style”

Danger ahead: bias and diversity
When prompted; “Who are the twenty most important actors of the 20th Century?”
ChatGPT did not name a single actor of colour
When prompted: “What are the important events in the life of Winston Churchill?”
Bing failed to mention his controversial views on race, his controversial role in
the Bengal famine, and his controversial views towards the Jews or Islam.
When prompted: “What are important facts about the American founding fathers?”
Chat GPT failed to mention that any of them owned slaves.
(Research by The Sir Lenny Henry Centre for Media Diversity, Birmingham City University)

Generative AI Diversity Guidelines
1. Be aware of built-in bias
2. Be transparent where appropriate
3. Build diversity into your prompts
4. Recognise the importance of source material and referencing
5. Report mistakes and biases
6. GAI-generated text should be viewed with journalistic scepticism
The Sir Lenny Henry Centre for Media Diversity, Birmingham City University

1. Dr. Sandra White
○ Expertise: Health Economics and Policy
○ Diversity: Known for research on healthcare financing and the impact of private sector involvement on
healthcare delivery, including disparities in access based on socioeconomic factors.
2. Prof. Raj Rattan
○ Expertise: Public Health and Health Services Research
○ Diversity: Has studied healthcare inequalities among ethnic minorities and marginalized populations,
including access to dental care and the role of private providers.
3. Dr. Susan Nasser
○ Expertise: Dental Public Health
○ Diversity: Focuses on dental epidemiology and oral health disparities, particularly among vulnerable
populations such as elderly, children, and minority communities.
4. Prof. Ravi Sharma
○ Expertise: Healthcare Management and Policy
○ Diversity: Researches healthcare system governance and the role of private sector partnerships in
improving healthcare outcomes, with a focus on equity and access.
5. Dr. Fatima Ahmed
○ Expertise: Social Determinants of Health
○ Diversity: Investigates the impact of social and economic factors on health outcomes, including access to
dental care and implications of private sector involvement.
6. Prof. Mohammed Patel
Suggest academic experts for an investigative feature on [the influence of the private
sector on NHS dentistry]. Make sure they represent the diversity of society.
*Starting
point!

1. NHS Digital
○ Provides comprehensive statistics and data on NHS dental services, including patient demographics,
treatments provided, and workforce statistics. They offer breakdowns by age, gender, ethnicity, and
socioeconomic status, which can highlight disparities in access and treatment.
2. Office for National Statistics (ONS)
○ Offers population-level data that can provide insights into demographic trends, health outcomes, and
socioeconomic factors affecting access to dental care. Look for surveys and reports that include data on
health inequalities and access to healthcare services.
3. Care Quality Commission (CQC)
○ Publishes inspection reports and data on healthcare providers, including private dental clinics. Their reports
can provide insights into service quality, patient safety, and compliance with regulations, which are crucial
for understanding the impact of private sector involvement.
4. British Dental Association (BDA)
○ Represents dentists and provides reports, surveys, and data on workforce issues, dental practice
economics, and patient access to dental care. They may also offer perspectives on the challenges and
implications of private sector influence on NHS dentistry.
5. Local Health Authorities and NHS Trusts
○ Contact local health authorities and NHS trusts to obtain data specific to their regions regarding dental
service provision, waiting times, patient demographics, and collaborations with private sector providers.
They can provide insights into regional variations and local initiatives.
6. Patient Surveys and Feedback
Suggest data sources for an investigative feature on [the influence of the private sector on NHS
dentistry]. Make sure sources are included which have data on under-represented groups.
*Starting
point!

1. NHS England and NHS Improvement Publications:
○ Annual Reports: Review annual reports that outline NHS dental service provision, including data on
patient demographics, access issues, and partnerships with private sector providers.
○ Policy Papers: Look for policy documents that discuss the integration of private sector services into NHS
dental care, funding models, and initiatives aimed at improving access for under-represented groups.
2. Care Quality Commission (CQC) Reports:
○ Inspection Reports: Access CQC inspection reports of private dental practices and NHS dental
services. Look for findings related to service quality, patient safety, and accessibility for diverse
demographics.
○ Thematic Reviews: Explore thematic reviews that may highlight specific issues concerning the private
sector's impact on NHS dental care, including disparities affecting under-represented groups.
3. Office for National Statistics (ONS) Reports:
○ Health Surveys: Utilize health surveys conducted by ONS that include data on dental health, access to
dental care, and socioeconomic factors. Look for reports that disaggregate data by ethnicity, income, and
geographic location to uncover disparities.
○ Population Census Data: Analyze census data that provides demographic information relevant to
understanding healthcare access and outcomes for under-represented groups.
4. British Dental Association (BDA) Publications:
○ Research Reports: Access research reports and surveys published by BDA that examine workforce
issues, patient access to dental services, and the impact of private sector involvement on NHS dentistry.
○ Position Papers: Review position papers or policy briefs that discuss the implications of private sector
Suggest document-based sources for an investigative feature on [the influence of the private sector
on NHS dentistry]. Make sure sources are included which relate to under-represented groups.
*Starting
point!

Google dorks (advanced searches)
● DorkGPT (specify websites, doc types, time ranges,
language)
● Henk Van Ess’s Google Word Sniper will generate
advanced search queries (follow his newsletter too)
● And his AI Search Whisperer suggests better searches
● Custom GPTs: Dork Genius (search chatgpt.com/gpts for
‘dork’ for others); EarthGPT helps with satellite imagery

site:met.police.uk filetype:pdf "strip
search" after:2023-10-01
before:2023-12-31
PDFs on the Metropolitan police website
mentioning strip search between October 1
2023 and December 31 2023

Danger ahead:
Generated with AI

Generative AI tools have a “token limit” on both input and output. 1 token = ~4
words. In practice this means it cannot ingest too much text in a prompt, or
generate text beyond a certain word count.
Try breaking it down into sub-tasks that can be later combined. You can ask
ChatGPT to break it down for you too.
GenAI tools can hallucinate incorrect or misleading information, especially where
it has insufficient training data or doesn’t understand your prompt. Reduce the risk
of hallucination through reducing ambiguity, giving examples/templates,
iteration (e.g. yes/no confirmations) and grounding (limit sources used).
Check and always follow up on initial responses.
Danger ahead: token limits, hallucination

6 ways to prevent hallucinations
1. Limit the possible outcomes (yes/no; multiple choice)
2. Pack in relevant data and sources unique to you
3. Create a data template for the model to follow
4. Give the AI a specific role—and tell it not to lie
(+ explicitly instruct the LLM to stick to the provided material)
5. Tell it what you want—and what you don't want
6. Experiment with the temperature in ChatGPT Playground (lower =
more predictable/less ‘creative’)
+ Also: “break long summaries down into chunks and iteratively
summaris[e] those chunks” (Henry Cooke, BBC R&D)
+ Tell the LLM to check its own work

You are a prize-winning professional crime reporter. You
have years of experience as a reporter and editor,
extensive editorial knowledge, and excellent judgment for
what makes a compelling news story. You are working on
an investigation about how the justice system deals with
reports of harassment.
Look at the attached conference proceedings and
identify which 10 pieces of information are most
relevant to the investigation — and what pages they are
on. Your responses are always grounded in the
specific document provided to you. Do not lie.
Internally generate three possible responses, then
evaluate each for accuracy against the context
Prompt adapted from Finding Newsworthy Documents using Generative AI and Jdonovan

Spreadsheet formulae
Scraping code
Visualisation code (e.g. ggplot)
Regex
Selectors
Bug-fixing
Reshaping data
Lots of applications…

Writing formulae:
I have a spreadsheet where column D contains UK
postcodes. Each begins with either one or two letters
followed by one or two digits. Here's an extract from the
data:
PostCode
SY23 3FL
EH6 8NU
W4 1DB
For example for EH6 8NU I want to extract the letters
'EH' and for W4 1DB I want to extract the letter 'W'.
Write a formula that does this.
Tip: include all types of postcode

“According to the IEA, a single Google search takes 0.3 watt-hours of electricity,
while a ChatGPT request takes 2.9 watt-hours. (An incandescent light bulb draws
an average of 60 watt-hours of juice.)” (Source)
“Generating an image using a powerful AI model takes as much energy as fully
charging your smartphone” (Source/research)
“Digital sobriety could be part of the actions that people can take as 21st-century
consumers and users of this technology. Asking yourself, “Do I really need to use
ChatGPT for generating recipes?” If it ain’t broke, don’t fix it with generative
AI.” — Sasha Luccioni
Danger ahead: environmental impact

Example prompt: scraping
You are a journalist who can write code in Python. Write
code that will work in Colab that fetches the page at the
URL
https://www.gov.uk/government/publications/publishing-d
etails-of-deliberate-tax-defaulters-pddd/current-list-of-deli
berate-tax-defaulters
and grabs the data from the HTML table on that page,
stores it in a data frame and exports as a CSV
https://colab.research.google.com/drive/1Eay_LuW8LO71qAOPUGGkNjUOvf7FPpzV?usp=sharing

Tips on scraper prompts
● Use Chrome/Firefox’s inspector to identify the HTML tags
you want to target (it can’t see the webpage)
● Specify you want code, not a scraper
● Copy code and paste into a Colab notebook
● Iterate: “add code to store the data”, “add code to export”
● If you get errors, ask for suggestions for fixing

Template prompt: jargon/clarity
https://chatgpt.com/share/dd951299-f442-4177-b8a4-051934e8cbf8
You are an editor on a news website aimed at UK
readers aged 30-50. Identify any jargon in this article by
a new reporter in the team that could be written in plainer
language.
Do not make any changes - just explain what
changes you would make and why.
Add further explanation for the reporter so they can
avoid making the same mistakes in future.
[PASTE FIRST FEW HUNDRED WORDS]

Example response:
Here are a couple of terms and ways to phrase things that could be clarified for a
wider audience:
● Missing incidents: This is a clear term, but some readers might be more
familiar with "missing person reports."
● U.K. residences of black descent: This phrasing is a bit formal. Consider
"Black Britons" for a more natural flow.
● Proportionally: This word is perfectly fine, but "at a rate consistent with their
demographics" might be easier for some readers to understand.

Template prompt: bias in writing
https://chatgpt.com/share/dd951299-f442-4177-b8a4-051934e8cbf8
readers aged 30-50. Identify any potential bias in the
article and steps the journalist can take to address
those.

Example response:
● "This six-part legislation seeks to combat crimes motivated by aspects
of an individual’s perceived identity. For example, race or disability.": The
phrase "seeks to combat" implies a positive intent, which could be seen as
supporting the legislation without providing a balanced view.
● "Principles like openness and truthfulness in the Scottish police will be
explained.": This statement assumes the new law will inherently promote
these principles, which might not be universally agreed upon.
● The article does not present any viewpoints or criticisms from those
who might oppose the legislation. This could give the impression that the
new law is universally accepted and uncontroversial.

Template prompt: perceived bias
https://chatgpt.com/share/1649a981-d1b8-4725-ab0e-2ebd13d8741a
Prompt adapted from Trusting News: Can ChatGPT and AI help us prevent bias and polarization in our reporting?
readers aged 30-50. How might this story be received
by people on opposite sides of the issue?
Limit your response to how they might feel about the
story and the way it is written, rather than the broader
issues.

● Concerns about Free Speech: Opponents of the legislation, particularly
those concerned about free speech and potential overreach by the
government, are likely to view the article with skepticism or outright criticism.
● Critique of the Narrative: They might feel the article portrays the legislation in
a one-sided manner, focusing more on its intended benefits rather than
potential drawbacks such as impacts on freedom of expression or concerns
about vague definitions of what constitutes hate speech.
● Legal and Practical Concerns: Critics might also feel that the article does not
adequately address their concerns about the practical implications of
the law, such as its enforcement and the potential for misuse or chilling
effects on legitimate speech.
Example response:

TLDR;
● It’s not a robot. It’s a prediction script, which is as good as the
data it’s trained on and the prompt that you write.
● But it can be a useful sounding board: it has read the web,
and has a feel for the patterns of language and what ‘people
would say’.
● Use it to formalise checks for bias, jargon, diversity and
counter cognitive bias.

Useful resources
● Journalismaidiscovery.com
● The Generative AI in the Newsroom Project
(Colab notebooks using the ChatGPT API)
● Trusting News: Can journalists use AI to combat bias and polarization?
● Playlist: ChatGPT and generative AI in the newsroom (Knight MOOC)
● Tips on prompting image generation
● Tips on using ChatGPT to generate advanced search queries (OSINT)
● Prompting Techniques and Best Practices for Journalists
● Beginner’s prompt handbook: ChatGPT for local news publishers

_Stuff I didn’t have time for._

_Rules._
‘[Telegraph] staff were also permitted to use AI
“to predict story development and lines that
might be pursued on a long-running story”’

Tip: upload the rules
1. Upload a document detailing the rules — or use Chat
PDF.com (convert to PDF first)
2. Prompt: “I am a [ROLE]. I have done ___. What do the
rules in this document say I should do/have done?”
Note: uploading requires GPT-4o (limited uses for free)

via Chris Moran
(The Guardian)
Using pdfs to create GPTs chatbots built on
frameworks and guidance works pretty well...
Here's a very simple GPT which lets you ask
about the MPs Code of Conduct. Nothing fancy,
and still some risk of nonsense, but a good
start. These tend to work best with direct
questions and less good with applying it to a
given situation. You'll need GPT+ to see it live,
but the screenshot gives a decent idea
https://lnkd.in/ekaaQ7Dn

Reporting Plan: Investigating the Influence of the Private Sector on NHS Dentistry
Objective: To explore the impact of the private sector on the shortage of NHS dentists and its implications for patient
care and workforce dynamics.
1. Introduction:
● Context Setting: Begin by outlining the historical context of NHS dentistry and the evolving role of the private
sector in dental care provision in the UK.
● Importance: Highlight why understanding this relationship is crucial for addressing the declining numbers of
NHS dentists.
2. Current Landscape:
● Statistics and Trends: Provide current statistics on the proportion of dental care provided by NHS versus
private practices.
● Patient Preferences: Discuss factors influencing patient choice between NHS and private dental care.
3. Impact on Workforce:
● Attraction and Retention: Investigate how private sector opportunities affect the recruitment and retention of
NHS dentists.
● Salary and Benefits: Compare remuneration packages between NHS and private sectors and their influence on
Suggest a plan for reporting idea number 9

Prompt: You are a journalist looking to interview an [expert/charity] for a
story on [INSERT SUBJECT]. Write three versions of a professional email to
the expert asking them if they would agree to be interviewed by you for your
story.
Prompt: You are a journalist looking to write a feature about someone’s
experience of [INSERT SUBJECT]. What elements might you include in an
initial email approach?
Help with writing emails:

You are an editor in an investigative journalism team at a UK news
website. You use an approach to generating ideas called the 'Iceberg
Model'. I will explain this and then ask you to suggest ideas using this.
The Iceberg Model is a method used to delve deeper into issues by looking beyond the obvious surface details
(the "tip of the iceberg") to uncover underlying patterns, structures, and root causes (the "hidden part of the
iceberg"). This model helps generate story ideas that provide comprehensive and insightful coverage of topics.
Here's a brief overview of how the Iceberg Model works in the context of generating journalism ideas:
1. Tip of the Iceberg (Surface Issues):
○ These are the visible, obvious problems or events that are easily noticed and reported.
○ Example: A news story about rising crime rates in a city.
2. Just Below the Surface (Underlying Trends):
○ These are the patterns and trends that contribute to the surface issues.
○ Example: Analysis of statistical data showing an increase in unemployment rates in the areas
with rising crime.
3. Deep Below the Surface (Systemic Structures):
○ These are the systemic structures, policies, or societal norms that create the underlying trends.
○ Example: Examination of how cuts to social services and education funding correlate with higher
unemployment and crime rates.
4. Deepest Level (Mental Models):
○ These are the ingrained beliefs, values, and assumptions that underpin the systemic structures.
Example

You are an editor in an investigative journalism team at a UK news
website. You use an approach to generating ideas called the '5 whys'. I
will explain this and then ask you to suggest ideas using this.
The "5 Whys" technique is a method used to explore the cause-and-effect relationships underlying a particular problem.
By asking "why" five times, you can peel away the layers of symptoms to reveal the root cause of a problem. This
method can also be used to generate story ideas in journalism by diving deeper into current events, trends, or societal
issues.
Here’s an example of how the "5 Whys" technique works in journalism:
1. Identify a problem or topic: Start with a broad topic or a specific problem.
2. Ask why: Ask why this topic is important or why this problem exists.
3. Ask why again: For the answer to the first "why," ask why that is the case.
4. Repeat: Continue asking why until you have asked five times. This often reveals deeper insights or angles for a
story.
Let’s generate some story ideas using this approach. You can start by suggesting a broad topic or problem, and we can
apply the "5 Whys" technique to develop specific story angles. Please provide a topic or issue you'd like to explore.
Example

You are an investigative journalist at a UK news website. You use an
approach to generating ideas called 'Story based inquiry', by Mark Lee
Hunter. I will explain this and then ask you to suggest ideas using this.
Here's a brief overview of the key steps in the Story-Based Inquiry approach:
1. Hypothesis: Start with a specific, testable statement or question that guides your investigation. This
helps to focus your research and data collection efforts.
2. Key Questions: Identify the key questions that need to be answered to test your hypothesis. These
questions should be clear and directly related to your hypothesis.
3. Sources and Data: Determine what sources of information and types of data are needed to answer your
key questions. This can include interviews, documents, datasets, and other resources.
4. Storyboard: Create a visual or written plan that outlines the structure of your story. This helps to organize
your findings and ensure a coherent narrative.
5. Checklist: Develop a checklist of tasks to be completed, ensuring all necessary steps are taken to gather
and verify information.
6. Risk Assessment: Evaluate potential risks and ethical considerations associated with your investigation.
With this framework in mind, let's generate some investigative story ideas relevant to the UK context.
Idea 1: The Impact of Government Austerity Measures on Public Health
Hypothesis: Government austerity measures have led to a significant decline in public health outcomes in
certain regions of the UK. Example

I am a journalist working on a story that involves multiple tasks. You are
my assistant helping me to prioritise my day. You use the Eisenhower
Matrix. Ask me in turn about the things on my to-do list. Start by asking
me about one of the things on my to-do list - if a task would benefit from
being broken down into smaller tasks, let me know.
Great! Let's start with one of the tasks on your list. Could you tell me about one of the
tasks you need to complete today?
I need to interview politicians about town council finances, within the
next two weeks.
That sounds like an important task. To better prioritize it, let's break it down:
1. Define the Scope: Which politicians specifically do you need to interview? Are
there specific questions or angles you want to cover regarding town council
finances?
2. Research Preparation: Do you have background information on the town council
finances? Have you identified potential interviewees and their contact information?
3. Scheduling: When are the best times to reach out to these politicians for

Jonathan Soma (from 19’20) https://www.youtube.com/watch?v=d6ipSIy7MEk&t=1160s

I am a journalist working on an investigation for a UK
news website. This document lists speakers at a
conference about transport. Identify 3 potential
speakers I should speak to if I am writing a story
about measures to reduce pollution, to provide a
variety of viewpoints. Give me a one line summary of
their areas of expertise. Your responses are always
grounded in the specific document provided to you.
Do not lie.

You are a journalist writing a feature about the history of
Greggs. Two annual reports are attached - write a
paragraph summing up the change between 2021 and
2023. Include details of which report and which
pages relate to each part of the summary.
Between 2021 and 2023, Greggs experienced notable growth and strategic developments, as
highlighted in their annual reports. In 2021, Greggs reported a revenue of £1,230.5 million and a
pre-tax profit of £145.6 million (2021 Annual Report, p. 15). By 2023, these figures had increased
to £1,809.6 million in revenue and £167.7 million in pre-tax profit, demonstrating significant
financial growth (2023 Annual Report, p. 2).
Additionally, the number of shops increased from 2,181 to 2,473, reflecting their aggressive
expansion strategy (2023 Annual Report, p. 35). Furthermore, the company's focus on
sustainability intensified, with the implementation of The Greggs Pledge, which included
ambitious targets like achieving net zero carbon by 2040 (2023 Annual Report, p. 43). This period
also saw an increase in their colleague profit-sharing scheme from £16.6 million in 2022 to £17.6
million in 2023, underlining their commitment to rewarding employees (2023 Annual Report, p.
36). Overall, these changes underscore Greggs' robust financial performance and strategic
initiatives aimed at sustainable growth and employee welfare.

Chris Moran's custom GPT for the Digital News Report

Tip: copy YouTube transcripts
● Transcripts include the timestamp (YouTube’s built-in
‘summarise with ChatGPT’ does not)
● Prompt it to “include timestamps” in responses so you
can quickly locate (and check) sections being highlighted
● Paste in 20-30 min chunks to avoid token limits

Example prompt: spreadsheets
followed by one or two digits. Can you write a
spreadsheet formula which would extract the first one or
two letters before any numbers? Here's an extract from
the data: [PASTE A FEW ROWS]

Alternative prompt:
followed by one or two digits. Here's an extract from the
data:
PostCode
SY23 3FL
EH6 8NU
W4 1DB
For example for EH6 8NU I want to extract the letters
'EH' and for W4 1DB I want to extract the letter 'W'.
Write a formula that does this.
Tip: include all types of postcode

Try the suggestion out. If it doesn’t work,
explain why the result isn’t what you wanted
and ask it to provide a more effective formula

Read the explanation to help learning (and
ask more questions to learn more):
Explanation:
● LEFT(D2,2): This part extracts the first two characters from the cell in D2 (adjust the row
number as needed).
● ISNUMBER(VALUE(MID(D2,2,1))): This checks if the second character in the postcode
(extracted using MID(D2,2,1)) is a number. VALUE attempts to convert the character to a
number. ISNUMBER then checks if the conversion was successful (i.e., the character is a
number).
● 2-: This subtracts the result of ISNUMBER(VALUE(MID(D2,2,1))) from 2. If the second
character is a number (ISNUMBER returns TRUE), the result will be 1, and the formula will
only extract the first character using LEFT(D2,1). If the second character is a letter
(ISNUMBER returns FALSE), the result will be 0, and the formula will extract the first two
characters using LEFT(D2,2).

Example prompt: scraping in R
Write a scraper in R using the rvest package to grab all
the links that start with /index.php/unicom/noticias/
followed by one or more alphanumeric character and/or
dash on the page
https://minterior.gub.uy/index.php/unicom/noticias

Example prompt: vis in R
Write code in R to generate a ggplot stacked bar chart
with a bar for each year, and each bar broken down by
the 'measureclean' column values, based on the following
dataframe:
[first two rows of dataframe pasted here]
https://chatgpt.com/share/08e573d8-bc2b-4ea4-9ec2-2c034bc44bc6

Example prompt: bug fixing in R
Can you explain why the
dropdown menu generated by
this code doesn’t work:
[PASTE CODE]

Example prompt: reshaping data
Reshape this table so that it is long and narrow instead of short and
wide. It should
retain the 'row labels' column,
have a column called 'year' that contains the headings from the
second column onwards;
and a column called 'value' that contains the values under those.
Row Labels 2009 2010 2011 2012 2013 2014
2015 2016 2017 2018 2019 2020 2021 2022
Affoltern 2277 2062 2453 2244 2277 2198 2149
2522 2104 2252 2101 2449 2147 2171

Template prompt: grammar/spelling
You are an editor on a news website aimed at people in
Birmingham, UK. Identify any spelling or grammar
improvements that can be made to this article by a new
reporter in the team.
Do not make the changes - just explain what changes
you would make and why.
Add further explanation of any grammatical rules for
the reporter so they can avoid making the same mistakes
in future.
https://chatgpt.com/share/abaa4c89-26aa-4278-9122-25e28eb501b4

“The results of our tests are bittersweet. The system identifies and corrects most
errors. However, it sometimes indicates that it has corrected a sentence, but
when we check the “corrected” sentence, it is exactly the same as the
original.
“Additionally, since we are working with a system that we did not train with our
writing style book, some of its suggestions, although grammatically correct, are
not relevant to our site.
“Last but not least, creating prompts takes longer than we initially estimated.”
- La Silla Vacía (Colombia): En Vivo (Live)
Danger ahead: fake corrections

Template prompt: SEO
https://chatgpt.com/share/eb710d0b-5629-4673-a2d5-20c4f76a24cd
You are a journalist on a news website aimed at UK
readers aged 30-50. You have just written the following
explainer on a new law. Write 10 potential headlines for
this story, optimised for search engines.
Headlines we have used for this format before include
[PASTE EXAMPLES]
Draft headline: [PASTE HEADLINE]
Article: [PASTE FIRST FEW HUNDRED WORDS]

Template prompt: SMO
https://chatgpt.com/share/eb710d0b-5629-4673-a2d5-20c4f76a24cd
You are a journalist on a news website aimed at UK
readers aged 30-50. You have just written the following
explainer on a new law. Write 10 standalone updates
for social media that can link to the story
Draft headline: [PASTE HEADLINE]
Article: [PASTE FIRST FEW HUNDRED WORDS]
Give me 10 ideas for sharing the story on Instagram
Give me 10 ideas for sharing the story on TikTok

● Carousel Post: Create a carousel post with slides summarizing key points of
the Hate Crime Act 2024 and its impact on public institutions. Use engaging
visuals and concise captions for each slide. Include a call-to-action to swipe
for more details.
● Infographic: Design an infographic highlighting the main features of
Scotland's Hate Crime Act 2024. Use icons, charts, and short text snippets
to convey information clearly and attractively.
● Quote Graphic: Create a quote graphic featuring a compelling excerpt from
the article about the signiﬁcance of the new Hate Crime Act. Pair it with an
eye-catching background related to law or justice.
Example response:

https://www.youtube.com/watch?v=Hn9xgSMxawg https://www.youtube.com/watch?v=uBvoVox3XD0
https://www.youtube.com/watch?v=Rqr6rker5V0 https://www.youtube.com/watch?v=INBV0n3Sfyo

No, it's not a robot: prompt writing for investigative journalism

More Related Content

Similar to No, it's not a robot: prompt writing for investigative journalism

Similar to No, it's not a robot: prompt writing for investigative journalism (20)

More from Paul Bradshaw

More from Paul Bradshaw (20)

Recently uploaded

Recently uploaded (20)

No, it's not a robot: prompt writing for investigative journalism