About & Methodology

On this page

• What is this?
• Who runs this
• Accountability Index
• Public inquiry tracking
• Coroner reports (PFD)
• AI assessment methodology
• Data freshness
• Editorial policy
• Limitations & caveats
• Corrections & updates
• Contact

What is this?

UK Inquiry Tracker brings together recommendations from public inquiries, tracking what was recommended, how the government responded, and what publicly available evidence exists of progress. It uses AI to search official sources and summarise the published evidence for each recommendation — these summaries are clearly labelled and should be treated as indicative, not authoritative.

It covers two main areas:

• Public Inquiries — 2645 recommendations tracked across 121 inquiries, with government responses and published evidence summaries
• Coroner Reports — Prevention of Future Deaths (PFD) reports issued by coroners under Regulation 28, tracking whether organisations respond within the legal 56-day deadline

Who runs this

This site is built and maintained by Matt Lewsey. Matt spent over 20 years in the UK Civil Service, most of it working on public inquiries and inquests. His roles included Director of the Grenfell Tower Inquiry, Director of the “Hillsborough Law” duty of candour, Secretary to the Hillsborough Independent Panel, Head of Inquiries and Inquests at the Ministry of Defence, and Head of the Cabinet Office’s central inquiries unit. He also led the design and initial setup of the UK Covid-19 Inquiry.

Why this exists

Public inquiries and coroners produce recommendations designed to prevent failures from happening again. But systematic follow-up is rare, and the data is scattered across multiple sources.

The government announced in February 2025 that it would establish a GOV.UK record of inquiry recommendations. This tracker brings together inquiry recommendations and coroner PFD reports independently, presenting the publicly available evidence so users can form their own view of progress.

Independence and funding

UK Inquiry Tracker is not affiliated with, endorsed by, or funded by the UK government, any political party, or any public inquiry.

For transparency: AI tools are used in the creation and maintenance of this tracker. Google Gemini generates automated evidence summaries for recommendations and PFD report summaries. Anthropic Claude assists with development. All AI-generated content is clearly labelled.

Contact

Get in touch | LinkedIn

Public inquiry tracking

Recommendations are catalogued using verbatim text from official inquiry reports. We do not summarise or paraphrase — the recommendation text on this site should match the source document word for word.

Responsible bodies are assigned based on who the inquiry directed each recommendation to, or which department the government's response assigns responsibility to.

What we track

Each recommendation is presented with two layers of information:

• Government Response — what the government said it would do:
  • Accepted Government accepted the recommendation
  • Accepted in Part Accepted in principle, or partially accepted
  • Not Accepted Government rejected the recommendation
  • Under Review Government is still considering its response
  • Awaiting Response No formal government response recorded
• Published Evidence — what official sources say has happened since:
  • AI-generated summaries of evidence found on gov.uk, legislation.gov.uk, and Hansard
  • Progress timeline entries attributed to named sources

A government may accept a recommendation without subsequently acting on it. This tracker presents the government's stated position alongside the publicly available evidence, so users can form their own view of progress.

Inquiry coverage

The tracker covers 121 UK public inquiries in three categories:

• Ongoing — inquiries that are currently sitting or have been announced but not yet reported
• Completed — inquiries that have published their final report. Where recommendations were made, these are catalogued with government responses and published evidence summaries
• Historic — older inquiries included for reference, often with limited recommendation tracking

Some completed inquiries have no recommendations catalogued because they made none (e.g. the Hutton Inquiry explicitly declined to make recommendations; the Dawn Sturgess Inquiry report contained no formal recommendations).

Coroner reports (Prevention of Future Deaths)

When a coroner believes action should be taken to prevent future deaths, they are required under Regulation 28 of the Coroners (Investigations) Regulations 2013 to issue a Prevention of Future Deaths (PFD) report. Organisations named in PFD reports have a legal duty to respond within 56 days.

We track PFD reports published by the Judiciary of England and Wales, including:

• Whether organisations responded within the legal deadline
• AI-generated summaries of coroners' concerns
• AI classification of responses (action taken, action planned, noted, or disputed)
• Trends across categories (e.g. hospital deaths, road traffic, suicide)

How we match PFD reports to organisations

PFD reports name addressees in free text (e.g. “the Chief Executive of University Hospitals Birmingham NHS Foundation Trust”). We use automated matching to link these to organisations in our database. This matching is not perfect — organisation names change over time, coroners use inconsistent naming, and some addressees are individuals rather than organisations. Where a match cannot be made confidently, the report is left unmatched rather than assigned incorrectly.

PFD response data should also be interpreted with caution: older reports (pre-2023) may not have responses uploaded to the Judiciary website, even where responses were submitted. Our response rate statistics use a two-year window to guard against less reliable older data.

Accountability Index (in development)

Beyond inquiry recommendations and coroner reports, we are building a broader Accountability Index that brings together data from across the oversight landscape. The aim is to provide a single view of how public bodies are meeting their obligations — linking inquiry recommendations with inspection findings, regulatory enforcement, ombudsman investigations, and parliamentary scrutiny.

The scorecard pages are available at accountabilityindex.uk.

AI assessment methodology

Each public inquiry recommendation has an automated evidence summary. These are generated by AI, which searches official sources and summarises the publicly available evidence relating to each recommendation. They should be treated as indicative, not authoritative.

Important: AI evidence summaries describe what has been published — they do not judge whether a recommendation has been “delivered” or “implemented”. The presence of published evidence does not prove real-world impact; the absence of published evidence does not prove inaction. Users should read the sources and form their own view.

Model

Assessments are generated using Google Gemini 2.5 Flash.

Evidence sources

For each inquiry, the system automatically searches four categories of official source:

• gov.uk — searches for “[inquiry name] recommendation implementation” and “[inquiry name] government response”
• Official government response documents — the formal government response URL, if one exists in our database
• legislation.gov.uk — searches for relevant legislation by inquiry name
• hansard.parliament.uk — searches for parliamentary debates mentioning the inquiry

Evidence is limited to official and regulatory sources only. The system is explicitly instructed not to cite NGOs, campaign groups, or pressure groups — even where their analysis may be accurate. This is a deliberate methodological choice to ensure summaries rest on verifiable official evidence.

What the AI is told

Recommendations are sent to Gemini in batches of 10, along with the searched evidence. The AI is instructed to:

• Summarise only what published sources say — not draw conclusions about whether a recommendation has been “implemented”
• Attribute every claim to a named source — using the format “According to [source]...”
• Include source URLs or traceable references — so users can verify the evidence themselves
• Never editorialise — report what sources say, nothing more
• Assess each recommendation individually — even within the same inquiry
• Write 2–3 sentences per recommendation

PFD AI summaries

Coroner report concerns and response classifications are also generated using Gemini 2.5 Flash. These summarise the coroner's concerns in 1–2 sentences and classify each organisation's response as:

• Action taken — the organisation has already made changes
• Action planned — the organisation commits to future changes
• Noted — the organisation acknowledged the report but described no specific action
• Disputed — the organisation disagreed with the coroner's findings

Known limitations of AI assessments

• We can only report what published sources say — we cannot independently verify whether a recommendation has actually been carried out in practice
• A government or public body saying it has done something is not the same as it being done — AI summaries may reflect official claims that have not been independently tested
• Assessments may lag behind recent government announcements
• The model can only assess based on evidence available on the four source websites listed above — it has no access to unpublished information, ministerial correspondence, or internal government tracking
• Absence of published evidence is not evidence of absence
• Batch processing means all recommendations for one inquiry share the same evidence context, which may cause the model to conflate related recommendations

Data freshness

Each inquiry shows a “last verified” date indicating when its data was last checked against official sources. These are displayed as freshness badges on the inquiry list page:

• Green — verified within the last 3 months
• Amber — verified 3–6 months ago
• Red — verified more than 6 months ago or never verified

We prioritise verifying inquiries with active government implementation programmes. Some historical inquiries with stable recommendations are verified less frequently.

Editorial policy

Evidence, not verdicts

This site does not claim that recommendations have been “implemented” or “delivered”. Implementation is complex, often partial, and the information needed to make that judgement definitively is rarely available from published sources alone.

Instead, we present:

• The government's stated position — accepted, rejected, or under review (a verifiable fact from official response documents)
• Published evidence of progress — legislation, policy changes, official progress reports, parliamentary scrutiny, and inspection findings
• Progress timeline — a chronological record of official updates, each attributed to its source

Users should read the evidence and form their own view. A government accepting a recommendation does not mean action has been taken. Legislation being passed does not automatically mean the recommended outcome has been achieved.

How conflicts between sources are handled

Where official sources disagree (e.g. a government progress report says “complete” but parliamentary scrutiny finds otherwise), we record both positions and flag the discrepancy.

Relationship to government data

In February 2025, the government announced it would establish a GOV.UK record of all public inquiry recommendations since 2024. This tracker is independent of that initiative and serves as an independent verification layer — comparing what the government says it has done with the available evidence.

Limitations & caveats

• This is an independent project with limited resources — coverage is not comprehensive
• Updates may lag behind official announcements, particularly for less prominent inquiries
• Published evidence of progress does not equate to full implementation — a recommendation may appear well-evidenced in official sources but poorly implemented in practice
• This is not an official government resource and should not be cited as one
• AI assessments are automated and have not all been manually verified
• Some recommendations are assigned to “UK Government” where the inquiry report did not name a specific department — this is accurate to the source but limits departmental accountability analysis

Corrections & updates

If you spot an error:

• Use the “Submit a Correction” button on any recommendation page to flag inaccurate data
• Or email us directly at the address below

Corrections are prioritised based on severity. Factual errors (wrong response status, wrong recommendation text) are corrected immediately. Disagreements about the strength of evidence are investigated and updated if additional sources support a change.

All significant changes are logged on the changelog.