Defra Content Audit Builder

Get a list of pages from Gov.uk to be included in your audit, with one reusable query

Start from every published page across any organisation, document type and topic, then use AI triage to narrow it down to the pages that really belong in your audit.

All of GOV.UK ~800k Keyword / phrase match 1,240 Organisation & type 320 Semantic include 150 Semantic exclude 94 Candidates 94 pages
Beta

A new service for policy and audit teams. Operated by a Defra supplier using data supplied by the GOV.UK Search API.

Start from a list of published live pages on GOV.UK

Define the set of pages you want once and generate a list of candidate pages for your audit or impact analysis.

You set the brief as a saved audit — a group of Filter Parameters — and you can re-run whenever policy moves.

Cast a wide keyword net

Every page you are looking for shares a keyword or phrase. Give the builder that list, and any page matching any one of them is included.

Be inclusive. You do not have to think of every term up front, and a broad net is cheap to widen.

Organisation and document type are there to narrow the scope and make the triage list manageable.

Refine later with semantic goals

Once the keyword net has gathered the candidates, you say what you actually mean in plain language: an inclusion goal for what belongs, and an exclusion goal for what to leave out.

The keywords decide what is considered; your semantic goals decide what is kept. Add and sharpen them after you have seen the candidates, with no need to re-run the search.

A two-pass AI applies your goals

The candidates go to a two-phase read against your goals. A generous first pass scores every page from 0 to 1 on the inclusion goal and keeps anything in scope. A strict second pass re-checks the keeps against the exclusion goal and can only remove.

Every decision comes with a score and a one-line reason in plain English, so you can review the candidate pages row by row and defend them.

From the whole of GOV.UK to an audit you can defend

Any match
Any page matching any keyword or phrase is included. Be inclusive; you do not need every term.
2 goals
One inclusion goal and one exclusion goal, added later, decide what is kept.
1 reason
A plain-English reason recorded for every page kept or dropped.

Cheaper because the easy pages never reach the model

Filters firstdeterministic, free, and auditable
AI lastonly on the pages that genuinely need judgement

The keyword net and the organisation and document-type scope remove most of the corpus with rules that cost nothing to run. The AI only reads the candidates that remain, so you spend budget where judgement is actually needed.

Every run is stamped with its model, token count and cost, so the price of an audit is never a mystery and two runs can be compared like for like.

Built for policy and audit teams

The Content Audit Builder is operated by a Defra supplier using data supplied by the GOV.UK Search API. It is a discovery and triage aid: it narrows the field and shows its working, so a person makes the final call.

Every audit shows how it was built, which filter removed what, and why the AI kept or dropped each page.