Google removes AI Overviews from liver test searches after Guardian probe

Google pares back some AI Overviews on liver test ranges after accuracy concerns

After questions about misleading health answers in AI Overviews, Google appears to have removed the feature for certain liver test range queries. An investigation reported that summaries presented single "normal" values without considering nationality, sex, ethnicity, age, assay, or lab-specific ranges.

Searches such as "what is the normal range for liver blood tests" and "what is the normal range for liver function tests" no longer show AI Overviews for many users. Variations like "lft reference range" reportedly did trigger summaries at first, but these also seem reduced now.

Google said it doesn't comment on individual removals and is focusing on "broad improvements." The company added that its clinical reviewers found many instances were supported by high-quality sources, though the feature behavior has changed since the initial reports.

Why this matters to clinicians

Reference ranges are context-dependent. They vary by population, instrumentation, units, and clinical setting. A single "normal range" in a generic summary can lead to false reassurance or unnecessary alarm for patients comparing their results at home.

Expect more patients to bring AI-generated answers into consults. Your best counterweight is clear, consistent education on local lab ranges and clinical context.

What seems to have changed

- Some liver test range queries no longer display AI Overviews.
- Related variants may intermittently show summaries, suggesting ongoing tuning rather than a blanket removal.
- News coverage of the removal now ranks high for these queries, which further suppresses the AI summary surface in practice.

Bottom line: don't rely on a stable experience. What appears in Search this week may shift next week.

Practical steps for your practice

• Standardize patient messaging: "Always use the reference range printed on your lab report; ranges online may not match our lab."
• Add a short explainer to patient portals and after-visit summaries on how ranges differ by lab and method.
• Give patients authoritative links you trust, and remind them these do not replace their clinician's interpretation.
• Train front-desk and nursing staff to triage "I saw online that my result is normal/abnormal" calls with a consistent script.
• For clinicians, document the lab method and local range when interpreting results in notes shared with patients.

Talking points you can use with patients

• "Different labs use different machines and units-so 'normal' isn't one number."
• "Your results need to be read in context: symptoms, medications, age, and your baseline."
• "If you search online, use it as background, not a diagnosis. Bring questions to us."

What advocates are saying

The British Liver Trust called the removal "excellent news," while warning it addresses a single set of queries rather than the broader issue of AI-generated health summaries. That concern aligns with what many clinicians see daily: helpful in parts, but unpredictable, and not patient-specific.

Helpful resources

• Coverage of AI Overviews and health query accuracy
• British Liver Trust: Liver function tests overview

Upskilling your team on AI literacy

If your clinicians or digital team need a quick, practical primer on evaluating AI outputs in care settings, explore concise learning paths built for organizational roles.

• AI Learning Path for Training & Development Managers
• AI Learning Path for Primary School Teachers

AI in search will keep shifting. Keep patient guidance simple, stick to local lab references, and make it easy for people to ask you before they act on an online summary.