Tens of Millions of Errors Per Hour: Investigation Reveals the 'Accuracy Illusion' of Google AI Search

marsbitPubblicato 2026-04-13Pubblicato ultima volta 2026-04-13

Introduzione

A New York Times investigation, in collaboration with AI startup Oumi, reveals significant accuracy and reliability issues with Google's AI Overviews search feature. Testing over 4,300 queries showed the accuracy rate improved from 85% (powered by Gemini 2) to 91% (Gemini 3). However, given Google's scale of ~5 trillion annual searches, this 9% error rate translates to nearly 57 million incorrect answers generated hourly. A critical finding is the prevalence of "unsubstantiated citations." For correct answers, the rate of citations that do not support the AI's summary surged from 37% to 56% with the Gemini 3 upgrade, making it difficult for users to verify information. The AI heavily relies on low-quality sources, with Facebook and Reddit being among its top-cited websites. Furthermore, the system is highly manipulable. A BBC journalist successfully "poisoned" it by publishing a fabricated article; Google's AI began presenting the false information as fact within 24 hours. Google disputed the study's methodology, criticizing its use of the SimpleQA benchmark and an AI model (Oumi's own) to evaluate another AI. The company maintains its AI Overviews, combined with its search ranking systems, perform better than the underlying model alone. Critics note this defense does little to bolster user confidence in the feature's reliability.

Author: Claude, Deep Tide TechFlow

Deep Tide Guide: A recent test conducted by The New York Times in collaboration with AI startup Oumi shows that the accuracy rate of Google Search's AI Overviews feature is approximately 91%. However, given Google's scale of processing 5 trillion searches annually, this translates to tens of millions of incorrect answers generated every hour. More troublingly, even when the answers are correct, over half of the cited links fail to support their conclusions.

Google is disseminating misinformation on an unprecedented scale, and most people are completely unaware.

According to The New York Times, AI startup Oumi, commissioned by the publication, used the industry-standard test SimpleQA, developed by OpenAI, to evaluate the accuracy of Google's AI Overviews feature. The test covered 4,326 search queries, conducted in two rounds: one in October last year (powered by Gemini 2) and another in February this year (upgraded to Gemini 3). The results showed that Gemini 2's accuracy was about 85%, which improved to 91% with Gemini 3.

91% sounds good, but it's a different story when considering Google's massive scale. Google processes approximately 5 trillion search queries annually. With a 9% error rate, AI Overviews generates over 57 million inaccurate answers per hour, nearly 1 million per minute.

Correct Answers, Wrong Sources

More alarming than the accuracy rate is the issue of "unsubstantiated citations."

Oumi's data shows that in the Gemini 2 era, 37% of correct answers had the problem of "unsubstantiated citations," meaning the links attached to the AI summary did not support the information provided. After upgrading to Gemini 3, this proportion increased instead of decreasing, jumping to 56%. In other words, while the model gives correct answers, it is increasingly failing to "show its work."

Oumi CEO Manos Koukoumidis pointedly questioned: "Even if the answer is correct, how do you know it's correct? How do you verify it?"

The heavy reliance on low-quality sources by AI Overviews exacerbates this problem. Oumi found that Facebook and Reddit are the second and fourth most cited sources for AI Overviews, respectively. In inaccurate answers, Facebook was cited 7% of the time, higher than the 5% rate in accurate answers.

BBC Journalist's Fake Article "Poisons" Results Within 24 Hours

Another serious flaw of AI Overviews is its susceptibility to manipulation.

A BBC journalist tested the system with a deliberately fabricated false article. In less than 24 hours, Google's AI Overview presented the false information from the article as fact to users.

This means anyone who understands how the system works could potentially "poison" AI search results by publishing false content and boosting its traffic. Google spokesperson Ned Adriance responded by stating that the search AI feature is built on the same ranking and security mechanisms used to block spam, and claimed that "most examples in the test are unrealistic queries that people wouldn't actually search for."

Google's Rebuttal: The Test Itself Is Flawed

Google raised several concerns about Oumi's study. A Google spokesperson called the research "seriously flawed," citing reasons including: the SimpleQA benchmark itself contains inaccurate information; Oumi used its own AI model, HallOumi, to judge another AI's performance, potentially introducing additional errors; and the test content does not reflect real user search behavior.

Google's internal tests also showed that when Gemini 3 operates independently outside the Google Search framework, it produces false outputs at a rate as high as 28%. However, Google emphasized that AI Overviews, leveraging the search ranking system, performs better in accuracy than the model alone.

Nevertheless, as PCMag pointed out in a logical paradox: If your defense is that "the report pointing out our AI's inaccuracies itself uses potentially inaccurate AI," this likely does not enhance user confidence in your product's accuracy.

Domande pertinenti

QWhat was the accuracy rate of Google's AI Overviews feature as tested by Oumi, and how many errors does this translate to per hour given Google's search volume?

AThe accuracy rate of Google's AI Overviews was found to be 91% in the test. Given Google's annual volume of 5 trillion searches, this 9% error rate translates to over 57 million inaccurate answers generated every hour.

QAccording to the Oumi study, what was the trend in 'unsubstantiated citations' between the Gemini 2 and Gemini 3 versions of the AI Overviews?

AThe problem of 'unsubstantiated citations' (where the provided links did not support the AI's answer) increased from 37% with Gemini 2 to 56% with the upgraded Gemini 3.

QWhich low-quality websites were identified as major sources frequently cited by Google's AI Overviews?

AFacebook and Reddit were identified as the second and fourth most frequently cited sources by the AI Overviews feature.

QHow did a BBC journalist demonstrate the vulnerability of Google's AI Overviews to manipulation?

AA BBC journalist tested the system by publishing a deliberately fabricated article. Within 24 hours, Google's AI Overviews began presenting the false information from that article as a factual answer to user queries.

QWhat were Google's main criticisms of the Oumi study's methodology?

AGoogle criticized the study for having 'serious flaws,' stating that the SimpleQA benchmark itself contains inaccuracies, that using Oumi's own AI model to judge another AI could introduce errors, and that the test queries did not reflect real user search behavior.

Letture associate

Two Legends Lost in Three Days: Is Google's AI Talent Dam Cracking?

In three days, Google lost two AI legends. On June 18, Noam Shazeer, co-author of the seminal "Attention is All You Need" paper and Gemini co-lead, left for OpenAI. Just 48 hours later, John Jumper, 2024 Nobel laureate and AlphaFold lead, departed DeepMind for Anthropic. This follows Andrej Karpathy joining Anthropic in May. These moves highlight a structural trend: top AI talent is concentrating at mission-driven, pre-IPO firms like OpenAI and Anthropic, while Google becomes a primary source. The exodus stems from a core mission mismatch. Google's ad-centric model often subordinates AI research to product and revenue goals, creating friction for pioneers like Shazeer, who returned in 2024 only to leave again. In contrast, OpenAI and Anthropic offer singular focus on pushing AI boundaries, whether towards AGI or safety-aligned models, which deeply appeals to top researchers like Jumper. Financial incentives amplify the pull. With both OpenAI and Anthropic nearing IPO, employees stand to gain immensely from equity, an upside Google's mature stock cannot match. Furthermore, the 2023 merger of Google Brain and DeepMind, intended to consolidate strength, has instead created cultural tension and slowed the path from research to product, as evidenced by Gemini's pace. This talent redistribution is reshaping the AI landscape. While Google retains vast data and compute resources, its true crisis is the quiet, continuous loss of the people who define the field's future. The real moat in AI is not infrastructure, but the concentration of brilliant minds—a battle Google is currently losing.

marsbit25 min fa

Two Legends Lost in Three Days: Is Google's AI Talent Dam Cracking?

marsbit25 min fa

Behind the AI Report Card, Lies a Chinese 'Exam Setter'

Beyond the familiar performance charts like MMLU-Pro and MMMU, which major AI models strive to ace, stands a key "examiner": Chinese-Canadian researcher Wenhu Chen. An assistant professor at the University of Waterloo and founder of TIGERLab, Chen addresses the crucial need for more rigorous AI evaluation. As models like GPT-4 began scoring near-perfect results on older benchmarks like MMLU, it became difficult to distinguish their true capabilities. In response, Chen introduced MMLU-Pro in 2024, featuring harder, more reasoning-focused questions with more answer choices, successfully reintroducing meaningful performance gaps. His work extends to multi-modal evaluation with MMMU and its enhanced version, MMMU-Pro. These benchmarks test a model's ability to understand and reason with complex information from images, charts, and text across diverse academic subjects, exposing the significant challenges even top models face in genuine comprehension. Chen's background in complex QA, table reasoning, and his experience at Google DeepMind on projects like Gemini inform his approach. He understands that effective benchmarks must anticipate how models might "cheat" by memorizing data or avoiding visual analysis. His lab also actively researches video understanding and generation models (e.g., UniVideo, Vamba), ensuring his evaluation work is grounded in practical model-building challenges. Now at Meta's Super Intelligence Lab, Chen continues his focus on multi-modal data and evaluation, representing the deep yet often unseen contributions of Chinese talent in shaping the fundamental tools of the AI industry.

marsbit37 min fa

Behind the AI Report Card, Lies a Chinese 'Exam Setter'

marsbit37 min fa

Alliance Co-founder's Letter to Entrepreneurs: Written at the Moment Cursor Sold for $600 Billion

Alliance Co-founder's Letter to Entrepreneurs: On Cursor's $60 Billion Sale Many aspiring founders see massive exits like Cursor's $60B sale and wonder why they can't achieve the same, often concluding opportunities are exhausted. But great companies aren't built in obvious, crowded spaces. Cursor, like Stripe, Figma, and Shopify before it, started with a non-consensus belief about the future. Before ChatGPT, they believed AI would transform knowledge work. They focused on a genuinely exciting domain, became their own customer, and obsessed over power users. Their journey involved years of "glass-chewing" effort before the market was ready. The pattern is consistent: identify a long-term technological shift, find a missed entry point, and execute for years before the trend becomes obvious. First-generation products (PayPal, Adobe, Amazon) prove a market exists. Second-generation winners (Stripe, Figma, Shopify) rebuild that market around new insights, technology, or changing customer behaviors. Founders must identify their phase in the cycle. Early entrants like Coinbase or Cursor focus on making new technology usable for power users. Later entrants find the "yin" to the established "yang"—the blind spots incumbents miss as they grow distant from individual users. The key is deep market immersion. Use every product in your space. Talk to users. Build an audience. Stop looking for ideas and start *seeing* them everywhere. Then, choose one. The idea must offer a 10x improvement or solve a "hair-on-fire" pain point—something severe enough that users are already crafting workarounds. When building, avoid feature bloat. Ask: why would someone switch? Great startups rarely force new behaviors; they improve familiar workflows with drastically lower friction (e.g., Cursor forked VS Code instead of creating a new editor). Distribution is the underestimated moat. Before product-market fit, achieve distribution-market fit. How do customers discover new tools? Founders like those at Airbnb, Stripe, and Cursor did unscalable, manual work to recruit early users. The final, unteachable ingredient is resilience. Cursor built for years pre-market, faced rejection, and persisted. So did Airbnb, Nvidia, and Rain (which launched post-FTX collapse). The lesson isn't that these founders were smarter, but that they stayed in the game long enough for their insights to compound. Framework: Spot technological cycles. Cultivate unique insight. Obsess over your market. Talk to customers. Find a hair-on-fire problem. Build the simplest wedge. Win your distribution channel. Above all, don't quit when it gets hard. Most people won't do these things consistently. The few who do build the next generation of great companies. Go build.

marsbit41 min fa

Alliance Co-founder's Letter to Entrepreneurs: Written at the Moment Cursor Sold for $600 Billion

marsbit41 min fa

Weekly Editor's Picks (0613-0619)

Weekly Editor's Picks (0613-0619): Market Insights & Analysis This weekly digest curates in-depth analysis often lost in the information flow, focusing on key insights across macro trends, investment, and technology. **Macro & Geopolitics:** With the Strait of Hormuz reopening and military conflict shifting to negotiation, markets are pivoting from "war shock" to "supply restoration." Trades include shorting crude risk premiums, longing airlines/tourism, Asian energy importers, and bond duration, while shorting inflation expectations. LNG, fertilizer, and chemical chains are also being repriced. **Investment & VC:** Ray Dalio advises against betting on concentrated AI giants dominating indices, advocating for diversified portfolios of high-quality, low-correlation assets instead. Analysis covers the 4-year crypto cycle, predicting the core surviving product by 2029 will be asset trading markets. Current BTC metrics suggest a potential bottoming zone, presenting a patient accumulation window. SpaceX's high-profile IPO at a $2.1T valuation faces scrutiny over fundamentals, with key watchpoints being its likely inclusion in the Nasdaq index and Q2 earnings. Concerns are raised about potential "gamma squeeze" and systemic risks if its narrative-driven valuation gets amplified by passive index funds. Robinhood (HOOD) is noted for breaking its high correlation with crypto, bolstered by its stock trading and new underwriting business. **Web3 & AI:** A warning highlights ~$1.8T in off-balance-sheet AI infrastructure commitments (purchase commitments, leases) as a potential systemic risk if AI monetization lags. AI models are being used for World Cup predictions, adding a new layer for betting markets. A cost breakdown of a $20 AI subscription reveals the supply chain from model companies to cloud, GPUs, and power. **Prediction Markets:** The emergence of prediction market "concept stocks" is noted, with Robinhood developing its own platform, Rothera, signaling a shift from market competition to a "channel war" for user access. **CeFi & DeFi:** The SpaceX IPO tested perpetual contract mechanisms for pre-IPO assets, highlighting challenges in handling corporate actions like stock splits on-chain. The de-pegging of STRC (Strategy's preferred share) to ~$89 reflects market concerns over MicroStrategy's capital structure and BTC-backed leverage model. BlackRock's covered-call Bitcoin ETF (BITA) offers yield but caps upside, appealing to yield-seeking institutions. **Ethereum:** An opinion piece argues Ethereum's core strength is its vast developer community and composability, solidifying its role as the default operating system for the financial internet. **Weekly Hot Topics:** Include the US-Iran deal reopening the Strait of Hormuz, Fed's hawkish hold, Anthropic restricting model access, SpaceX acquiring Cursor, and a humorous stock surge for "Liuliumei" due to its "LLM" ticker.

marsbit46 min fa

Weekly Editor's Picks (0613-0619)

marsbit46 min fa

Alliance's Co-Founder's Letter to Entrepreneurs: Written on the Occasion of Cursor's $60 Billion Sale

In this letter to entrepreneurs, Alliance reflects on the success of Cursor's $60 billion sale to Elon Musk, using it as a case study to counter the misconception that opportunities in crowded fields like AI or crypto are exhausted. The piece argues that great companies like Cursor, Stripe, Figma, and Shopify are not built by geniuses with perfect ideas, but by founders who start with a non-consensus belief about the future and build for years before that future becomes obvious to everyone. They identify long-term shifts, find overlooked entry points, and execute relentlessly. The framework for success involves: 1. **Identifying your place in the technology cycle**: Early-stage opportunities focus on making new tech usable for power users (e.g., Coinbase, Cursor). Later-stage opportunities involve finding the "yin" to an existing "yang"—the blind spots of first-generation players (e.g., Stripe vs. PayPal, Figma vs. Adobe). 2. **Cultivating unique insights**: Immerse yourself deeply in the market. Use every product, talk to users, and build an audience. Insights will emerge naturally from deep engagement. 3. **Finding a "hair-on-fire" problem**: Look for a 10x improvement or a severe, urgent pain point. The strongest signal is people already building clumsy workarounds. 4. **Building a focused MVP**: Don't just add features because you can. Ask why users would abandon their current tool for yours. The best startups rarely force new behaviors; they improve familiar workflows with drastically lower friction. 5. **Winning a distribution channel**: Distribution is often the moat. Before product-market fit, achieve channel-market fit. Find where your customers are and build an engine to reach them, even through unscalable, manual efforts initially. 6. **Persistence**: The final, unteachable ingredient is resilience. Success stories like Cursor, Airbnb, and Nvidia involved years of grinding, rejection, and perseverance when the path forward seemed unclear. The conclusion is that there is no secret. Most people fail to consistently execute these steps over the long term. The few who do build the companies that define the next era. The world is yours to create.

链捕手51 min fa

Alliance's Co-Founder's Letter to Entrepreneurs: Written on the Occasion of Cursor's $60 Billion Sale

链捕手51 min fa

Trading

Spot
Futures
活动图片