TL;DR:

• UK government content designers are warning that AI overviews from Google and other search engines are scraping outdated GOV.UK pages, producing confident but wrong answers to citizen queries.
• One example: a Google AI overview cited the cost of charity setup as £13 online or £40 by post — figures from an unmaintained legacy page — against the actual current cost of £100 online or £124 by post.
• The Department for Business and Trade has audited 150 stale pages meeting specific criteria (not updated in five years, fewer than 11 views in that period, no active owner) and redirected them; DBT is also trialling six-monthly review cycles with visible review dates.

AI overviews from Google and other search engines are confidently presenting obsolete UK government information as fact by scraping unmaintained GOV.UK pages, according to senior content designers at the Department for Business and Trade — a pattern that directly undermines trust in UK public services.

How the failure mode works

A Google search for the cost of setting up a charity in the UK returned an AI overview citing £13 online or £40 by post for Companies House incorporation. Both figures were pulled from an outdated, unmaintained page. The actual current cost is £100 online or £124 by post. When The Register tried the same query, Google’s AI overview first told them incorporation was free, then offered “roughly £13-£183+” the next day. Neither was right.

Senior content designer Giorgio Di Tunno and content operations lead Neil Starr set out the structural problem in a GOV.UK blog post: “In the past, most of those outdated, niche pages would fall into the 0-view abyss, never to be stumbled on again.” AI summarisation has collapsed that safety mechanism. Pages published by defunct departments — the Department for Business, Energy and Industrial Strategy was specifically flagged — remain crawlable and now get pulled into answers shown to millions. DBT’s blunt framing: “The problem isn’t that the government is trying to trip people up (it isn’t), but that inconsistent information surfaced by the AI overview feels that way to users.”

What DBT is doing about it

DBT audited GOV.UK pages matching four criteria: not updated in five years, fewer than 11 views in that period, purporting to carry current information, and having no active owner. The result was 150 pages, now redirected to archived copies, current GOV.UK equivalents or relevant legislation. The department is also testing a six-monthly review cycle with “last reviewed” and “next review” dates visible at the bottom of each page — a change content designers say has gone down “extremely well with real users” because it establishes an explicit freshness signal.

The Department for Education’s head of design Mark Edwards has warned separately that AI summaries produce “misleadingly narrow or incomplete answers” to government-services questions. Edwards’ framing points to the wider design challenge: “We now need to design with the expectation that much of what we publish will be read indirectly, atomized, summarized or reinterpreted by systems we don’t control.”

Looking Forward

For UK public services communications more generally, the DBT findings have direct procurement and design implications. Departments will need to formalise page-ownership and review cycles at much greater depth than historical practice, because the cost of stale information is no longer near-zero visits but actively wrong answers surfaced to millions. Expect GOV.UK’s design system to evolve with explicit review-date disclosure, and for AI-summarisation testing to become a standard content-quality check before publication. The harder problem — how to design content that survives reinterpretation by systems no one controls — is not yet solved anywhere.