TL;DR
Google removed certain AI Overviews for health queries after a Guardian investigation found the feature returned inaccurate and potentially harmful information. The company disabled specific liver-test queries but left other flagged answers accessible while pointing to internal clinician reviews and high-ranked sources.
What happened
A Guardian investigation uncovered that Google’s generative AI feature, AI Overviews, produced misleading health information at the top of search results. After experts raised alarms, Google disabled particular queries — including searches about normal ranges for liver blood tests — but did not remove all the examples cited. The investigation found the feature sometimes returned raw numeric tables naming enzymes such as ALT, AST and alkaline phosphatase without vital context or adjustments for age, sex or ethnicity. Experts warned that these oversimplified summaries could falsely reassure patients with serious liver disease and lead them to forgo follow-up care. The Guardian also noted a separate error about pancreatic cancer guidance recommending avoidance of high-fat foods, which contradicts standard clinical advice. Google declined to comment on the specific removals to The Guardian and told The Verge that internal clinicians review content and that many summaries are supported by high-quality sites.
Why it matters
- AI summaries shown at the top of search results can strongly influence user decisions, especially around health.
- Presenting raw lab ranges without demographic context risks misinforming patients about their condition.
- Design choices that summarize top-ranked pages can propagate SEO-driven inaccuracies into authoritative-seeming answers.
Key facts
- The Guardian reported that Google’s AI Overviews returned inaccurate health information in top search results.
- Google disabled specific queries related to 'normal range for liver blood tests' after experts flagged the results.
- AI Overviews produced raw tables listing enzymes such as ALT, AST and alkaline phosphatase without contextual guidance.
- The feature did not adjust reported ranges for patient demographics like age, sex or ethnicity, per the report.
- Experts warned the summaries could falsely reassure people with serious liver disease and delay care.
- The Guardian also highlighted an instance where the AI suggested pancreatic cancer patients avoid high-fat foods, contradicting typical clinical guidance.
- Vanessa Hebditch of the British Liver Trust said interpreting liver function tests is complex and requires more than comparing numbers.
- Google declined to comment on specific removals to The Guardian; to The Verge it said clinicians review summaries and that many are supported by reputable sources.
- The problem traces to a design that summarizes highly ranked web pages, which can include SEO-gamed or unreliable content.
- Some flagged AI Overviews remain accessible via query variations, according to the Guardian, and Google says the feature appears only where it has 'high confidence.'
What to watch next
- Whether Google will remove additional problematic AI Overviews identified by the Guardian or other experts (not confirmed in the source).
- If Google changes how AI Overviews incorporates demographic context and clinical caveats for lab results (not confirmed in the source).
- Any announcements about expanded external audits, independent reviews, or regulatory scrutiny of AI Overviews (not confirmed in the source).
Quick glossary
- AI Overviews: A Google feature that generates short summaries of web content using a large language model and top-ranked search results.
- Liver function tests (LFTs): A panel of blood tests that measure enzymes and proteins to help assess liver health; interpretation depends on clinical context.
- ALT / AST: Enzymes commonly measured in liver blood tests; abnormal levels can indicate liver injury but require clinical interpretation.
- SEO (Search Engine Optimization): Practices aimed at improving a webpage’s ranking in search results; high rank does not guarantee medical accuracy.
Reader FAQ
Which queries did Google disable?
Google disabled specific liver-test queries such as searches about the normal range for liver blood tests, according to the report.
Did Google comment on the removals?
Google declined to comment to The Guardian on the specific removals and told The Verge that clinicians review AI Overviews and that many summaries are supported by reputable sites.
Are other problematic AI Overviews still live?
Yes. The Guardian found that slight variations of the original queries still produced AI Overviews, and some other examples flagged earlier remained accessible.
Will Google fix the underlying issue with AI Overviews?
not confirmed in the source
UNSAFE AT ANY SEED Google removes some AI health summaries after investigation finds “dangerous” flaws AI Overviews provided false liver test information experts called alarming. BENJ EDWARDS – JAN 12,…
Sources
- Google removes AI health summaries after investigation finds dangerous flaws
- 'Dangerous and alarming': Google removes some of its AI …
- Google removes multiple AI health summaries as experts …
- Google removes some health-related questions from its AI …
Related posts
- Google Chrome 151 to End Support for macOS 12 Monterey, Warning Issued
- WorkJourney: AI-powered private work journal for employees and managers
- Fabrice Bellard’s ts_zip: LLM-based text compressor promising high compression ratios