When AI summarizes advice about medicine, money or the law, mistakes carry real consequences. Search engines and answer tools treat these topics differently from recipes or travel tips, and publishers in these fields face a higher bar. This article explains what happened when that bar was tested, and how to build content that earns trust and survives being excerpted by AI SEO.
A Case That Changed the Conversation
In January 2026, a Guardian investigation examined Google’s AI Overviews for health queries. According to reports, Google removed AI Overviews for searches asking about the normal range for liver blood tests and liver function tests. The problem was missing context. The summaries listed raw enzyme figures without essential context, which risked leading seriously ill patients to wrongly think they were healthy.
Google disputed part of the criticism. A spokesperson said an internal team of clinicians found that in many cases the information was not inaccurate and was supported by high-quality websites. The story also didn’t end there, since variations on those questions, such as “lft reference range,” still generated AI summaries. propakistanieuronews
For publishers, the lesson is not about one query. It’s that your page can be summarized in a way you didn’t intend, and the safest defense is writing content that holds up when its numbers are pulled out of context.
Why These Topics Face Higher Standards
Google’s quality framework calls such subjects YMYL, short for “Your Money or Your Life.” One summary describes them as topics that can affect health, financial stability, safety or well-being, with heightened E-E-A-T expectations. E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness. It is not a direct ranking factor but the quality framework behind Google’s ranking systems, and the rater guidelines call trust its most important element.
Exposure is also high. Conductor’s 2026 benchmarks report, cited by Search Engine Land, found that healthcare content appears most frequently in Google’s AI Overviews, followed by financial services. These are exactly the categories where being quoted badly does the most harm.
The Trust Chain
Many guides reduce E-E-A-T to “add an author bio.” A more useful approach is to treat trust as a chain with five links, where weakness in any one undermines the rest. This is a working framework built from the evidence above, not an official standard.
- Who Wrote It
Name a real author with relevant credentials, and link to a profile page that explains their background. iPullRank advises publishers to ensure authors have visible credentials and a digital footprint, such as professional bios, that AI systems can verify. Don’t invent credentials, and don’t use stock-photo “doctors.”
- Who Checked It
A writer and a qualified reviewer are different roles. For medical pages that might be a licensed clinician. For tax or investment content, a certified professional. For legal content, a practicing attorney. State the reviewer’s name, qualification and review date on the page, and publish a short editorial policy explaining how reviews happen.
- What Evidence Supports It
Cite primary sources: regulators, peer-reviewed research, professional bodies and public agencies. Describe what the source actually says in your own words and name it clearly. Avoid vague phrases such as “studies show.”
- When It Was Last Verified
Show a visible “last reviewed” date and keep it honest. Update the date only when someone has actually rechecked the content. Rules, rates and clinical guidance change, and stale advice is a risk, not just an SEO weakness.
- What Context Travels With It
This is the link most publishers miss, and it comes straight from the liver-test case. Write so that key figures never appear without their conditions. If you publish a reference range, put the caveats in the same sentence or the next: that ranges vary by laboratory, age and sex, and that a result should be interpreted by a clinician. An excerpt that carries its caveat with it is safer for readers and harder to misuse.
Write for Worldwide Readers, Not Just One Country
Standards for health, tax, law and finance differ by country. Coverage of the Google removals in South Africa noted that there are no local safeguards ensuring AI summaries reflect regional clinical guidelines or population-specific reference ranges. If your audience spans many markets, say which jurisdiction your advice applies to, and build separate pages where rules truly differ. A page that says “in the United Kingdom” or “under Indian tax law” is clearer to readers and to machines than one that pretends to be universal. capeargus
Use AI Carefully in Your Own Process
AI tools can help draft, summarize and organize, but in these fields a human expert must own the final claim. As Search Engine Land’s guidance for regulated industries puts it, teams should use AI appropriately with mandatory human oversight. Never publish unreviewed AI output on a topic where a wrong number could hurt someone.
An Audit You Can Run This Month
- List every page that touches health, money, safety or law.
- Check each for a named author, a named reviewer, cited primary sources and a real review date.
- Search each page’s key numbers and ask: if this figure appeared alone in an AI summary, would it mislead?
- Add caveats beside any figure that fails that test.
- Test five core questions in Google and two other AI tools, and note whether your pages are cited and how they’re described.
Sarkar SEO builds this kind of audit into its work for clients in sensitive categories, because trust problems are cheaper to fix before a page gains reach.
Realistic Expectations
None of this guarantees citations. AI systems still select sources in ways no one fully understands, and results can shift. What you can control is the quality and clarity of what you publish. Content that is accurate, reviewed, current and properly caveated is more likely to stand up to both algorithms and scrutiny.
