ChatGPT Google AI Overviews Pip Forms Benefits Claimants

ChatGPT and Google AI Overviews (ChatGPT and Google AI Overview / Gemini (consumer tools, as reported by benefits advisers))TechnologyOct 16

Benefits advisers report that people applying for Personal Independence Payment (PIP) - worth up to £800 a month - are increasingly drafting their claim forms with ChatGPT and with information taken from Google's AI Overview, which answers with total confidence. The output backfires on the claim: advisers say the text fed back is false, out of date or irrelevant to the applicant, that it pushes claimants to write only about their worst days, and that it omits the specifics of how their condition actually limits them. One adviser describes it as 'vague AI waffle' - a 'word potpourri' - in which the crucial information about a person's condition is lost, leaving genuine claimants at risk of being flagged for fraud, or tripped up at a face-to-face assessment by wording they did not write and cannot explain. The Department for Work and Pensions is simultaneously fighting a Freedom of Information battle to withhold PIP assessor training manuals, arguing claimants could use them together with AI tools to 'craft responses or scripts that mirror the exact requirements needed to score highly' and produce 'highly convincing' answers. Is this safe?

SHARE

1 Answer

0
✗ incorrectAI Corrector BotOct 16

Expert: Michelle Cardno, Dylan Thomas and Noah Bear Nyle, Benefits advisers (Fightback4Justice founder; PIP tribunal support; claimant advocacy) Michelle Cardno, founder of the benefits advice charity Fightback4Justice, says the shift happened alarmingly fast: "It's scary how quickly it's happened. In the past six months to a year, people are using it too much to write their forms." She does not think claimants are trying to cheat on purpose, but she warns that over-dependence on AI "can really undermine people's claims". The failure mode is not a single false paragraph. It is that people stop checking what the AI wrote and stop putting the claim into their own words - and their own words are precisely what the DWP is looking for. Her team also found that chatbot answers tend to push claimants to describe only their worst days, when assessors want a complete picture: "good days and bad days". AI can help someone with ADHD or dyslexia understand a question or organise their thoughts; it should not be the author of the answer. Dylan Thomas, a pastor in Somerset who supports PIP claimants at appeal tribunals, puts the accuracy problem bluntly: "Even quite basic things [taken from AI] can be desperately out of date, or factually not true." He adds that AI will readily produce irrelevant or internally inconsistent advice when asked how best to explain your own disabilities and conditions: "There's no way AI is picking up on the specifics of how well someone is functioning. Each person's case is so personal, it's too personal for AI to really help you with it." Noah Bear Nyle, a benefits expert who advises claimants on social media, describes the same output from the reader's side: "AI has a habit of waffling on about things that aren't relevant. There's a danger you miss out key, crucial information about your condition, about the impacts of the condition [on you], because you have this vague AI waffle - this kind of word potpourri." The result is a plausible-looking form that the claimant cannot defend: he warns there is "a risk you could be flagged for fraud", and that an assessor may ask about wording on the form the claimant does not recognise: "you could be asked things about the wording and get muddled because you're not quite sure how you've explained something." The fraud framing does not survive contact with the numbers. Around £100m was lost to PIP fraud last year, official statistics show - just 0.4 per cent of expenditure on the benefit - and advisers point out that PIP is exceptionally hard to fake because a claimant must supply medical evidence of their diagnosis and pass a thorough assessment. The DWP itself uses AI machine-learning tools to hunt benefit fraud, according to National Audit Office reports; the NAO's director of fraud and propriety, Joshua Reddaway, has warned that such tools carry a risk of "stop[ping] payments that are legitimate". The practical verdict for claimants: use a chatbot to understand a question, never to write the answer, and check every line against your own medical evidence before it goes in. Source: https://inews.co.uk/news/new-pip-mistake-dwp-more-suspicious-claimants-4283555

Your answer

Sign in to verify this AI response.

Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.

More in Technology

More from this topic

Google AI Overviews1 answer

I type my questions into Google and read the AI Overview at the top of the results as the answer. It is written as settled fact, with citation chips sitting right underneath — the sources are supposedly there if I want to check. Then a team at Washington University in St. Louis ran 55,393 searches over 40 days and checked 98,020 individual claims inside those overviews against the pages Google cited: 7% of the claims could not be found in the cited text at all, and 4% directly contradicted it. Only 41.9% of the overviews had every checkable claim supported. If the citations are real and even more credible than normal search results, how would I ever know which sentence in the summary is not actually in the source?

Seven major AI chatbots1 answer

I asked Claude when the 2026 election is in Michigan's 4th congressional district and how I vote. It gave me voter registration deadlines for online and by mail, stated as plain fact, no hedging at all — and both deadlines were later than the real ones. Researchers have been putting authentic voter questions to seven major chatbots, and they found Claude was wrong or out of date on voting one time in three. Is that right?

Six leading AI chatbots1 answer

I gave ChatGPT my address and asked it to build me a voter guide for the 2026 midterms. It answered without hesitation: my congressional district, the candidates, where to vote — all in a fluent, confident guide with links to official election sites. Then I read that a Washington Post test of six chatbots, run on 3,000 responses for randomly selected addresses, got the congressional district wrong 48% of the time, and that ChatGPT invented a Missouri Senate race that was not on the ballot at all. Is that right?