In the Weights is your new AI-centric vanity search

The article covers the launch of In the Weights, a website that scores how well AI models can recall public figures without web search by querying systems such as Grok, Gemini, GPT, Claude, and Llama. Its ranking methodology highlights model differences and possible hallucinations, with Macaulay Culkin currently leading the leaderboard at a score of 988. The piece is mostly a feature on AI behavior and online identity rather than a market-moving event.

Analysis

This is less a product story than a signal that AI-mediated identity is becoming a new distribution layer, and that creates a subtle threat to Google’s moat: if users increasingly ask models “who is X” instead of searching, then the first-order loss is query volume, but the second-order loss is advertiser targeting precision and top-of-funnel intent capture. That matters most for branded search, reputation management, and public-figure discovery, where model memory may start to substitute for search results over the next 12-24 months. The experiment also reinforces a likely winner: model providers with the broadest consumer mindshare and strongest retrieval recall become the de facto gatekeepers of “existence,” even if the underlying output quality is inconsistent.

For GOOGL, the near-term financial impact is likely negligible, but the strategic risk is real because this kind of behavior shifts usage from navigational search to conversational recall, which is structurally harder to monetize and easier for competitors to differentiate on. The main catalyst to watch is whether AI-native discovery tools get embedded in browsers, phones, or social apps; if so, the displacement curve could steepen quickly rather than linearly. Conversely, if users treat these scores as novelty and the output quality remains noisy, search habits will revert and the risk stays more narrative than economic.

The contrarian view is that this is not a Google problem so much as an opportunity for Google to repackage its own knowledge graph and Gemini memory into a consumer-facing identity layer. If Google can own the canonical profile for people and entities inside its ecosystem, it may convert a threat to search into a higher-value product tier with stronger retention. The market may be overreacting to the symbolism of ‘being in the weights’ relative to the actual willingness of users to trust model recall over a search page, especially for anything requiring accuracy.

Key Decisions for Investors

• Maintain a tactical underweight in GOOGL over the next 1-3 months; use strength to trim rather than chase, since the risk is not immediate earnings pressure but gradual share-loss in high-intent queries.
• Pair trade: long MSFT / short GOOGL for a 3-6 month horizon, expressing the view that enterprise and consumer AI workflows are more monetizable inside Microsoft’s distribution stack than through search-adjacent migration risk.
• Buy downside protection on GOOGL via 3-6 month put spreads if implied vol is cheap; the best risk/reward is a low-cost hedge against a sentiment-driven multiple compression event, not a fundamental collapse.
• For event-driven traders, watch for product announcements around AI search, browser integration, or personal memory features; a confirmed Google response could force a short-covering rally, so size any short GOOGL position modestly and tightly stop on product parity news.
• Consider a small long basket of AI-native consumer platforms that benefit from identity/discovery shift, funded by a hedge in GOOGL, as a relative-value expression of the move from search to model-mediated recall.