Blood Cancer UK reported that page views on its leukaemia information page fell 53 percent over the preceding twelve months, while page views in the charity's own online community forum rose 86 percent and engagement rose 143 percent. PPC Land published the figures on 17 August 2026 in reporting on UK charities losing search traffic to Google AI answers.
These are the charity's own reported numbers, not the output of a controlled study. No comparison group was published, no method was described for isolating AI answers from every other cause of a traffic change, and nothing in the reporting establishes that the people who stopped reading the disease pages are the same people who turned up in the forum. What the figures do give is a shape, and the shape is consistent across six pages in the same organisation.
What Blood Cancer UK reported
The declines are not uniform across the charity's content. They are steepest on the page covering the most searched condition and shallower on the rarer ones. Leukaemia lost the most at 53 percent. MGUS fell 45 percent, myeloma 32 percent, lymphoma 28 percent, the AML and MDS page 20 percent, and MPN 17 percent.
Kate Keightley of Blood Cancer UK, quoted in the reporting, described a growing appetite for lived experience, saying that while AI can give an instant summary it cannot replace the reassurance that comes from speaking to somebody who has been through the same thing. The charity also stressed that newly diagnosed people need information that is accurate, trusted and appropriate, which is the point that makes this more than a traffic story.
The page-level figures
Every figure below is as reported by PPC Land, covering the twelve months before publication. Each row is one page on the Blood Cancer UK site.
| Blood Cancer UK page | Reported change in page views |
|---|---|
| Leukaemia | Down 53 percent |
| MGUS | Down 45 percent |
| Myeloma | Down 32 percent |
| Lymphoma | Down 28 percent |
| AML and MDS | Down 20 percent |
The MPN page, down 17 percent, is the sixth in the set and the smallest decline reported.
The forum moved the other way
Over the same twelve months, the charity's online community forum recorded an 86 percent rise in page views and a 143 percent rise in engagement. That is the detail that separates this report from the general run of publisher traffic complaints. Two content types on the same domain, under the same brand, in the same period, moved in opposite directions.
A plausible mechanism sits in the difference between the two formats. A page that defines a condition and lists its symptoms answers a question that has one correct answer, which is precisely the kind of answer a generative system can assemble and present without a click. A thread in which somebody describes what their third round of treatment felt like has no canonical version to summarise. It is testimony rather than reference, and it is attached to a person the reader wants to hear from directly.
That reading is inference, not something the source proves. The reporting does not test causation, and other explanations exist: a redesign, a change in internal linking, a shift in how the charity promotes its forum, or ordinary year-on-year variation. It is worth stating the alternatives rather than letting a tidy story stand unchallenged.
Other organisations in the same reporting
Blood Cancer UK is not alone in the PPC Land piece. Five other UK charities reported declines. The Brain Tumour Charity reported a site-wide drop of about 27 percent, with donations reported as stable. Mind and the Mental Health Foundation are also named, with Mind reported as saying 60 percent of people now turn to chatbots for mental health support. Save the Children recorded a 303 percent year-on-year increase in AI Overview appearances while its clicks and impressions fell, which is the same pattern in a different sector: more surface area inside the answer, less traffic out of it.
The commercial side of health publishing appears in the same reporting through Ziff Davis, which wrote off 54.8 million US dollars against health media assets as advertising revenue fell 6 percent. A write-off of that size is an accounting judgement about future earnings, so it is a statement about what the company expects health media to be worth, not only about what happened last quarter.
The opt-out exists and it is blunt
Google began testing a Search Console toggle for opting out of AI answers on 3 June 2026, under UK regulatory pressure. The control available now works at domain level. Page-level control is not scheduled until 3 March 2027. In between, the CMA Publisher Conduct Requirement obligations take legal effect on 3 December 2026.
The gap between those two dates is the practical problem for a health publisher. A charity that wants its symptom pages excluded from AI summaries while keeping its fundraising and campaign pages eligible cannot do that yet. The choice on the table is the whole domain or nothing, which for most organisations is not a choice at all. Anyone weighing it should also note that opting out removes the AI answer, not the ranking behaviour around it, and the reporting gives no data on what traffic looks like after a domain opts out.
What these figures are not
Health content is a category where being roughly right is not good enough, so the limits deserve to be spelled out rather than buried.
- These are figures reported by named organisations about their own properties. They are not a study, and no methodology was published alongside them.
- No causal test was run. AI Overviews are the explanation offered in the reporting, and the numbers are consistent with it, but consistency is not proof.
- The sample is a handful of UK charities. Nothing here supports a claim about health publishers in general, in the UK or anywhere else.
- No Thai data appears in the reporting at all. Any read across to Thailand is reasoning about a mechanism, not evidence about a market.
- Page views are the metric reported. Conversions, donations, helpline contacts and appointment bookings mostly are not, and The Brain Tumour Charity's stable donations against a 27 percent traffic drop is a direct warning against assuming the two move together.
- This article is about publishing and search visibility. It is not medical guidance, and none of it should be used to decide where to get health information.
What a health publisher can check
The checks below follow from the pattern in the figures rather than from any assumption about what caused it.
- Split your health content by format, not by topic. Definitional and symptom pages on one side, patient stories, clinician commentary, Q and A threads and anything with a named human attached on the other. Compare the two trends over the last twelve months.
- Look at impressions against clicks per page in Search Console. Impressions holding while clicks fall is the Save the Children pattern, and it looks different from a page that simply lost rankings.
- Check whether the pages losing traffic have a named, credentialed author and a review date. Where a reader does land, that is what distinguishes your page from a summary.
- Track something downstream of the pageview: enquiries, bookings, newsletter signups, helpline volume. A traffic drop with stable outcomes is a different problem from a traffic drop with falling outcomes, and only one of them justifies emergency spending.
- Find out whether your organisation would want the domain-level opt-out if it had to decide today, and write down why. Page-level control is not due until 3 March 2027, so the decision is all or nothing until then.
Structuring content for a category where accuracy and authorship carry regulatory weight is the core of healthcare marketing, and the question of which formats survive being summarised sits inside generative engine optimisation.
What this means for Thai marketers
Thai hospitals, clinics, pharmaceutical brands and insurers publish a great deal of exactly the content type that dropped hardest in this reporting: pages that define a condition, list its symptoms and explain a treatment in general terms. If the mechanism behind these figures holds, that library is the exposed part of the estate, and it is often the part that took the most budget to build.
The source says nothing about Thailand and contains no Thai data. What can be reasoned, carefully, is that the content which resisted the decline in this one dataset was first-person and community content, and that a Thai equivalent would be patient stories in Thai, commentary attributed to a named doctor with visible credentials, and moderated question and answer threads. Those formats are harder to produce than another symptom page and harder to summarise away.
Two cautions specific to the Thai market. Medical advertising here is regulated, and patient testimonial content carries compliance requirements that a UK charity forum does not face, so the format cannot simply be copied across. And a community forum only works if somebody moderates it, which is an ongoing staffing commitment rather than a content project with an end date. An unmoderated health forum is a liability, not an asset.
The wider planning point is that a Thai health brand measuring content performance on pageviews alone will read this shift as a failure of SEO. It may instead be a change in which formats earn a visit at all, in which case the answer is a different content mix rather than more of the same pages. That question belongs in content strategy before it belongs in a technical audit.
FAQ
Did Google AI Overviews cause Blood Cancer UK's traffic drop?
The reporting attributes the decline to AI answers, but no causal test was published. What exists is a set of figures from the charity showing disease-specific information pages falling while its forum grew, over the same twelve months. That is consistent with the explanation and does not prove it.
Which health pages lost the most traffic?
The leukaemia page, down 53 percent, was the steepest decline reported by Blood Cancer UK, followed by MGUS at 45 percent, myeloma at 32 percent, lymphoma at 28 percent, AML and MDS at 20 percent and MPN at 17 percent.
Can a publisher stop Google using its pages in AI answers?
Only at whole-domain level for now. Google began testing a Search Console opt-out toggle on 3 June 2026 and page-level control is not scheduled until 3 March 2027, so a publisher cannot currently exclude some pages and keep others eligible.
Does this apply to health publishers in Thailand?
The source did not say. It reports on UK charities and contains no Thai data, so any application to Thailand is reasoning about a mechanism rather than a measured result.
Should a health site stop publishing symptom and definition pages?
Nothing in the reporting supports that, and for many organisations those pages still serve people who need them. The figures argue for adding formats that a summary cannot replace, not for deleting reference content that a smaller number of readers still reach.
Where this leaves the work
One charity's twelve months is not a rule, and it is the clearest published example so far of which health formats a generative answer absorbs and which it leaves alone. If your organisation publishes health content in Thailand and wants to work out which parts of it are exposed, the Relevant Audience health marketing service page sets out how that assessment is approached.
Source: PPC Land.
