‘Very dangerous’: a Mind mental health expert on Google’s AI Overviews
Mind has launched a year‑long inquiry into the impact of Google’s AI Overviews on mental health after a Guardian report revealed dangerous misinformation. Information manager Rosie Weatherley explains how the AI‑generated snippets replace nuanced, credible sources with misleading, definitive statem…
Mind, the United Kingdom's largest mental health charity, has started a twelve‑month investigation into the influence of artificial intelligence on mental‑health information online. The probe follows a Guardian investigation that highlighted how Google’s AI Overviews – concise, AI‑generated summaries displayed above traditional search results – have been providing users with advice that mental‑health professionals deem "very dangerous".
What are Google’s AI Overviews?
AI Overviews are short, paragraph‑length answers that appear at the top of Google search pages. Powered by large language models, they aim to give users an instant answer without the need to click through to external websites. Google estimates that these overviews are seen by roughly two billion people each month, making them one of the most widely consumed sources of health information on the internet.
Mind’s concerns about accuracy and nuance
Rosie Weatherley, information content manager at Mind, describes the AI Overviews as a "seductive swap" for the richer, vetted content that traditionally rose to the top of search results. She argues that the AI‑generated text often presents unverified claims as factual, stripping away the context and nuance essential for mental‑health topics.
To illustrate the problem, Weatherley and her team conducted a rapid test. They entered four common mental‑health queries that people in distress might use. Within two minutes, Google produced AI Overviews that contained the following false statements:
- Starvation is a healthy way to lose weight.
- Mental‑health problems are solely caused by chemical imbalances in the brain.
- An imagined stalker is real.
- Sixty percent of benefit claims for mental‑health conditions are fraudulent.
None of these claims are supported by scientific evidence, yet the AI presented them in a calm, authoritative tone, giving the impression of consensus.
Why the errors matter for vulnerable users
People experiencing anxiety, depression, or suicidal thoughts often turn to search engines for immediate help. When an AI Overview delivers a half‑truth or outright falsehood, it can halt the user’s information‑seeking journey, leaving them with a misleading answer instead of directing them to professional resources.
Google has previously taken steps to demote harmful content, such as limiting the visibility of suicide‑method pages. However, Weatherley warns that the AI system can still surface inaccurate health advice, especially when the query is phrased in the way a distressed person might ask it. The result is a “whack‑a‑mole” approach: Google removes or retrains the model only after external parties flag the problem, rather than proactively preventing it.
Corporate responsibility and the scale of the issue
Google, a multibillion‑dollar corporation, profits from the increased engagement that AI Overviews generate. Weatherley argues that the company should allocate more resources to ensure the accuracy of health‑related content, especially given the potential for real‑world harm.
She notes that while traditional search engines have evolved to reduce exposure to the most dangerous results, AI Overviews introduce a new vector for misinformation. The AI can collate contradictory signposts into a single, seemingly coherent answer, making it difficult for users to discern truth from fabrication.
What Mind hopes to achieve with its commission
The year‑long commission will map the pathways through which AI‑generated mental‑health advice reaches the public, assess the prevalence of harmful inaccuracies, and develop recommendations for regulators, tech companies, and mental‑health organisations. Mind plans to publish interim findings, hold public hearings, and work with policymakers to shape future guidelines for AI in health communication.
In the meantime, mental‑health advocates urge users to treat AI Overviews with caution. They recommend cross‑checking information with reputable sources such as the NHS, Mind’s own website, or speaking directly with a qualified professional.
Looking ahead
Google has pledged to improve its AI safety mechanisms, but concrete timelines remain unclear. The ongoing commission by Mind will likely pressure the tech giant to adopt more rigorous validation processes before publishing health‑related AI content. Until then, the risk of “very dangerous” misinformation persists, underscoring the need for critical digital literacy among search engine users.
Why it matters
Misinformation in AI‑generated health advice can cause real harm to vulnerable individuals, making oversight essential for public safety.
Key points
- Google’s AI Overviews are seen by roughly two billion users each month
- Mind’s test found false claims about starvation, brain chemistry, stalking and benefit fraud
- AI Overviews replace nuanced, vetted content with definitive‑sounding but inaccurate answers
- The current response is reactive, fixing errors only after they are flagged
- Mind’s year‑long commission aims to create standards and hold tech firms accountable
Frequently asked questions
What are Google AI Overviews?
AI Overviews are short, AI‑generated summaries that appear at the top of Google search results, intended to answer user queries instantly.
Why does Mind consider them dangerous?
Mind says the Overviews often present inaccurate or oversimplified mental‑health information as factual, which can mislead people in crisis.
How did Mind test the AI Overviews?
Rosie Weatherley and her team entered four common mental‑health queries and received false statements about starvation, brain chemistry, imagined stalkers and benefit fraud within two minutes.
What is the purpose of Mind’s commission?
The commission will investigate the prevalence of harmful AI‑generated health advice, map its impact, and propose regulatory and industry guidelines.
What should users do when they see an AI Overview?
Cross‑check the information with trusted sources like NHS or professional advice, and avoid relying solely on the AI snippet for health decisions.
