Google's AI answers are pushing health charity websites down the search results. Their patient forums are growing anyway.

A Google search for "what to do when I get blood cancer" now returns a 270-word AI-generated summary above the scroll, with Blood Cancer UK's expert page pushed below the fold. Twelve months on, the charity's leukaemia information pages have lost 53% of their traffic, according to figures shared with Press Gazette.

The traffic did not vanish. It migrated within the same site. Blood Cancer UK's online community forum, where patients and families talk to other patients, gained 86% in page views and 143% in engagement over the same period. Cancer UK, Mind, the Mental Health Foundation, The Brain Tumour Charity and Save the Children also told Press Gazette their web traffic has fallen significantly over the past year, with the same pattern: factual reference pages down, peer-support content holding or growing.

Google AI Overviews, the AI-generated summaries that now sit above the standard list of blue links, are designed for narrow factual queries: definitions, symptoms, what-to-do-next questions. That is the shape of a frightened, first-person health search. The summary above the fold synthesises content from across the web; the charity whose clinicians wrote most of that material lands further down, where it is far less likely to be clicked. The disclaimer under the AI box reads "AI generated responses may include mistakes," and sits one tap away from a decision a patient or carer may make that day.

Blood Cancer UK measured the page-level breakdown. Leukaemia information pages fell 53% in twelve months. MGUS (monoclonal gammopathy of undetermined significance, a precancerous blood condition) was down 45%, myeloma 32%, lymphoma 28%, acute myeloid leukaemia and myelodysplastic syndromes 20%, myeloproliferative neoplasms 17%. The general blood cancer hub held flat. The pattern is consistent: the more the query fits a single AI-summarisable answer, the bigger the drop.

"The pattern is so consistent that we can see it in our data: pages answering narrow factual questions are the ones being hit," Kate Keightley, associate director at Blood Cancer UK, told Press Gazette. "What is particularly interesting is the growing appetite for lived experience, for the ability to speak to someone who has been through it. AI can give you information. It cannot sit with you while you wait for your test results."

Keightley framed the shift as a re-pricing rather than a fight. As Google's answer engine has moved AI summaries above charity links, encyclopaedic, factual pages, the kind a charity can produce at scale and a search engine can summarise in a paragraph, are losing the search result. Peer community, lived experience, condition-specific reassurance from someone who has had the diagnosis, are the parts AI cannot synthesise, and those parts are growing on the same domains.

Health queries are the worst case for an answer engine that admits it can be wrong. The "AI generated responses may include mistakes" disclaimer is now adjacent to authoritative guidance on what to do at the moment of a new diagnosis, or when a loved one is being told the next step. A patient who reads the summary and stops scrolling is, for the first time in the search era, a default state of search design rather than an edge case.

Encyclopaedic, factual pages still have their purpose. The front door has moved. The peer-support community, the forum thread where someone whose leukaemia was diagnosed last year answers the question a frightened reader typed at midnight, was always the part of the charity web that a search engine could not commoditise. It is now the part the search engine is sending more people toward.

The price of a factual answer has fallen to near zero. The price of being told, by someone who has been where you are, what the next week actually feels like, has not.