AI Search Signal #1 — March 2026

April 1, 2026

This is the first edition of the GEOlikeaPro Signal - my curated intelligence brief on the AI search changes that actually move e-commerce brand visibility. Every item links to the primary source. No speculation, no filler. If I can't point you at the source, it doesn't go in here.

🤖 Google upgrades AI Overviews to Gemini 3 globally

Gemini 3 is now the default model for AI Overviews worldwide. Users can ask follow-up questions directly from an AI Overview and jump into AI Mode for a full conversational search. AI Overviews are used by over 1 billion people.

For your store: Those follow-up questions are the part people miss. A surface-level mention gets you into the first answer and then dropped on the follow-up. From what I've seen, the pages that survive the second and third question are the ones with real depth - structured product data and FAQ schema. Build for the follow-up, not just the first hit.

Source: blog.google

🔬 Google publishes TurboQuant - Cloudflare CEO calls it "Google's DeepSeek moment"

Google Research published TurboQuant (ICLR 2026), a compression algorithm that shrinks AI inference memory by 6x with zero accuracy loss. Cloudflare CEO Matthew Prince posted: "This is Google's DeepSeek moment. There is so much more room to optimize AI inference for speed, memory usage, power consumption, and multi-tenant utilization." Memory chip stocks dropped within hours - Micron -3%, Western Digital -4.7%, SK Hynix -7.3% over two days.

For your store: Read this one as a cost signal. Cheaper inference means more AI search queries at lower cost, which means this channel keeps growing whether you're ready or not. The pie is getting bigger - the only question is whether you're on the plate.

Source: research.google

🕷️ Cloudflare data: AI crawlers send 50B requests/day - 80% is training, not search

Cloudflare reports AI crawlers generate 50 billion+ requests daily across their network. 80% is for model training, 18% for search, 2% for user actions. GPTBot is the most blocked bot (312 domains) but also the most explicitly allowed (61 domains). OpenAI's crawl-to-referral ratio sits at 1,700:1 - they crawl 1,700 pages for every 1 click they send back.

For your store: That 1,700:1 ratio is the whole argument. Here's what I'd actually do: allow the search-specific crawlers (OAI-SearchBot, ChatGPT-User) and block the training one (GPTBot). You keep your ChatGPT Search visibility and stop donating your catalog to model training you get nothing back from.

Source: blog.cloudflare.com

🕷️ Cloudflare catches Perplexity bypassing robots.txt with stealth crawlers

Cloudflare documented Perplexity deploying undeclared crawlers that rotate IPs and spoof real-browser user agents to access blocked content. Cloudflare delisted Perplexity as a verified bot and reported a 400% rise in robots.txt bypass rates between Q2 and Q4 2025. PerplexityBot's request volume grew 157,490% year-over-year despite its tiny 0.2% share.

For your store: Don't kid yourself that robots.txt is a wall - against Perplexity it's a suggestion. If you genuinely need a page off-limits, enforce it with WAF rules or Cloudflare's AI Crawl Control. A directive Perplexity ignores is not protection.

Source: blog.cloudflare.com

🔍 Perplexity ships Comet iOS browser + Computer for Enterprise

Perplexity launched Comet, an AI-native browser for iOS, and expanded Computer to all Pro and Enterprise subscribers - routing tasks across 20 specialized models with 400+ app integrations. Also released pplx-embed-v1 search embedding models at 0.6B and 4B parameter scales.

For your store: Comet turns Perplexity from a search engine into the browser itself. People will research and buy without ever leaving it. The takeaway I keep coming back to: your product pages have to work for a human reader and an AI extractor at the same time, on the same page. You don't get to pick one audience anymore.

Source: perplexity.ai/changelog

🤖 Claude web search hits GA - free code execution with search

Anthropic's web search tool is now generally available (no beta header required). Code execution is free when paired with web search or web fetch. Claude Sonnet 4.6 delivers improved agentic search performance while consuming fewer tokens.

For your store: This is the quiet one, but it matters. Any Claude API developer can now build a shopping agent that hits the live web by default. If ClaudeBot can crawl your product pages and they carry structured data, you show up in those agent workflows for free. If it can't, you're invisible to a channel that just turned on.

Source: docs.anthropic.com

🤖 OpenAI starts testing ads in ChatGPT - paid tiers stay ad-free

OpenAI launched its first advertising test on February 9, 2026, for free and Go tier users in the US. From their official blog: "Ads do not influence the answers ChatGPT gives you. Answers are optimized based on what's most helpful to you. Ads are always separate and clearly labeled." Plus, Pro, Business, and Enterprise tiers remain ad-free. OpenAI also announced Shopify partnerships for in-chat checkout.

For your store: "Separate from organic recommendations" - for now. I've watched this movie before with Google. Pair the ads test with the Shopify in-chat checkout and what you're really looking at is organic visibility in ChatGPT turning into a direct revenue channel. Earn the organic mention while it's still free to earn.

Source: openai.com

📊 Google fires March 2026 core update + spam update in the same week

Google released the March 2026 core update (March 27) and spam update (March 24) simultaneously. The core update affects search results globally with a two-week rollout. Core updates directly impact which pages are selected as sources for AI Overviews.

For your store: Here's the link people don't draw: if your organic rankings move over the next two weeks, your AI Overview citations move with them. They're not two dashboards anymore - they're one. Watch them together or you'll misread both.

Source: searchenginejournal.com

Analysis: what this actually means for e-commerce GEO

The crawl landscape is splitting. OpenAI runs three separate crawlers with different robots.txt behavior. Perplexity bypasses blocks entirely. Google feeds Gemini from its Shopping Graph (50B+ product listings, 2B hourly updates). Every engine sources product data differently, so I'll say it plainly: there is no single "optimize for AI" move anymore. You manage visibility per engine, or you don't manage it at all.

Inference is getting cheaper. TurboQuant cuts AI memory 6x. DeepSeek already proved efficiency gains accelerate adoption. Cheaper inference means more AI queries, which means AI search keeps growing as a channel. Adam Dorfman, AVP of Product Management at Birdeye, put it directly: "Brands can no longer afford to ignore GEO. AI engines now decide what people see."

The organic window is narrowing. ChatGPT and Gemini are both bringing in ads. The brands that build AI visibility now, while organic citations are still free, will have the edge when paid placement becomes the default. I've seen this exact parallel play out in early Google SEO: the brands that ranked organically before AdWords went on to own the early paid landscape too. Same story, new engine.

GEOlikeaPro tracks your brand's visibility across ChatGPT, Perplexity, Claude, and Gemini in a single dashboard. See where you stand.