AI Search Signal #2 — April 2026
Signal #2 of the GEOlikeaPro briefing - what moved in AI search this past week and what it actually means for your store. Every item links to a primary source. If I can't point you at the source, it's not in here.
🛒 Shopify opens ChatGPT to millions of merchants by default
Shopify's Agentic Storefronts program is live. Merchants get out-of-the-box access to ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini - managed centrally from the Shopify Admin, no apps, no separate integrations, no transaction fees beyond standard processing. A new Agentic Plan also surfaces non-Shopify merchants' products via Shopify's tools in ChatGPT and Gemini.
Stay in the loop
Get news and updates about GEO, AI search and new features. Unsubscribe anytime.
Is your brand a Ghost or a Guide on AI?
See if AI knows your brand. We ask Gemini and Claude live - in ~5 seconds, no signup.
If you're on Shopify, your products are now discoverable inside ChatGPT by default - you didn't opt in, the platform did it for you. Go check your Shopify Admin for the Agentic Storefronts toggle, and verify your Product schema is complete, because ChatGPT reads your JSON-LD before it reads anything else. "Discoverable by default" only helps if the data behind it is there.
Source: shopify.com/news
🕷️ Cloudflare AI Crawl Control hits GA - plus Redirects for AI Training
Cloudflare's Agents Week (April 13-17) moved AI Crawl Control from beta to general availability and introduced Redirects for AI Training - a feature that automatically converts <link rel="canonical"> tags into HTTP 301 redirects for verified training crawlers (GPTBot, ClaudeBot, Bytespider). The toggle enforces your canonical hierarchy for crawlers without touching humans or search engines. Customizable 402 response codes are now open to all paid Cloudflare customers, not just pay-per-crawl users.
If you've been fighting AI crawlers slurping up deprecated URLs - old product pages, abandoned variants - turn on Redirects for AI Training. It tells GPTBot and friends to respect your canonicals, which saves crawl budget and keeps outdated content from surfacing in AI answers as if it were current. This is one of the few enforcement levers that actually has teeth.
Source: cloudflare.com/agents-week
🔍 Perplexity hits $500M ARR - Computer for Mac rolls out + $3M Billion Dollar Build
Perplexity's revenue jumped 50% in a single month, hitting $500M annualized in April. It now serves over 100 million monthly active users. Personal Computer for Mac launched to Perplexity Max subscribers ($200/mo). They also announced the Billion Dollar Build - an 8-week competition offering up to $3M ($1M seed + $1M compute + $1M upside) to founders building on Perplexity Computer. Registration opens April 14, finale June 9.
Read the direction here: Perplexity is moving from search engine to AI agent platform. Personal Computer means the user doesn't browse your site at all - they ask an agent to research, compare, and buy. Your product pages have to serve structured data to that agent, not just render pretty HTML for a human who's no longer the one looking.
Source: pymnts.com
🧠 Anthropic ships Claude Opus 4.7 + Claude Design + Claude Mythos Preview
Anthropic had a loaded week. Claude Opus 4.7 went generally available (April) with stronger software engineering, better vision, sharper instruction following, and more reliable long-running agent work. Claude Design (April 17) launched as a research preview - generate prototypes, slides, and one-pagers from plain-English prompts, powered by Opus 4.7. And Claude Mythos Preview (April 7) launched Project Glasswing, a frontier model tuned for computer security tasks.
Opus 4.7 sharpens Claude's agentic performance - which means Claude's web search and shopping agents got more reliable, and they're evaluating your store. Anthropic also changed its crawler model earlier this year. Make sure your robots.txt allows Claude-SearchBot (search index) and Claude-User (live queries), not the deprecated anthropic-ai - if that's still your only Anthropic directive, you're invisible and don't know it.
Source: anthropic.com/news
🤖 OpenAI raises $122B, launches $100 Pro plan, expands ads to AU/NZ/CA
OpenAI announced a $122B funding round to accelerate the next phase. On April 9 a new $100/month ChatGPT Pro tier launched between Plus ($20) and Pro ($200), with 5x more Codex usage than Plus. A major Codex update pushed it beyond coding into computer use, web workflows, image generation, and automations. Ads rolled out to Free and Go users in Australia, New Zealand, and Canada (previously US-only).
ChatGPT ads are still separate from organic answers - but watch the geographic rollout pace, that's the real signal. Paid tiers stay ad-free. If you sell in AU/NZ/CA, ChatGPT ads are now a buyable channel. Organic visibility is still free, but I've seen this movie with Google: the free window narrows, it doesn't stay open.
Source: openai.com
💬 Google rolls out Gmail AI Overviews + Gemma 4 open models
Gemini 3 now powers AI Overviews in Gmail - summarizing long threads into concise answers, with the ability to ask questions of your inbox (Pro and Ultra subscribers). On April 2, Google released the Gemma 4 family of open-weight models (gemma-4-26b-a4b-it and gemma-4-31b-it), available on AI Studio and through the Gemini API for developers needing local or fine-tuned deployments.
More Gemini 3 surfaces means more AI answers synthesized from your content. And Gmail AI Overviews means your promotional and order-confirmation emails get summarized too. If transactional email is a primary brand touchpoint for you, its content quality now feeds AI visibility - that's a surface most stores have never optimized once.
Source: blog.google
📊 Cloudflare: AI training crawlers ignore canonicals by default
The case for Redirects for AI Training came straight from Cloudflare's own data: training crawlers routinely ingest deprecated URLs - old product pages, variants, discontinued items - because they read content at crawl time and ignore rel="canonical" signals entirely. That stale content then surfaces in AI answers as if it were current.
If you've ever renamed or consolidated product URLs, your old versions may still be feeding AI responses right now. Two fixes: (1) enable Cloudflare's Redirects for AI Training, or (2) serve explicit 301s for deprecated URLs so crawlers physically can't reach the stale content. Don't assume the canonical tag is doing the job - the data says it isn't.
Source: blog.cloudflare.com
Analysis: three shifts worth watching
With Agentic Storefronts rolling out broadly, "being invisible to ChatGPT" stops being a niche worry and becomes an everyone problem. The merchants who optimize Product schema and allow OAI-SearchBot are the candidates AI agents pick from. Everyone else gets filtered out at the data layer, quietly, with no error message.
Cloudflare's new tools - Redirects for AI Training, customizable 402s, AI Crawl Control GA - mark the end of "just add robots.txt" as an enforcement strategy. The infrastructure layer is becoming the control point. If your site is on Cloudflare, your AI visibility now depends on Cloudflare settings as much as on your HTML, and the two can silently disagree.
Perplexity's pivot to Personal Computer, Anthropic's Opus 4.7 agentic gains, OpenAI's Codex computer-use update - all three labs are pouring money into agents that do tasks for users, not just answer questions. For e-commerce that means the shopper of 2027 is an AI making the decision on a human's behalf. Your product page has to convince the agent, not just the person. Start writing for that reader now.
GEOlikeaPro tracks your brand's visibility across ChatGPT, Perplexity, Claude, and Gemini in a single dashboard. Sign up free to see where you stand.
FAQ
What is Shopify Agentic Storefronts and do I need to do anything?
Agentic Storefronts is Shopify's default integration that makes every Shopify merchant's catalog discoverable in ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini. It's enabled by default for most merchants — check your Shopify Admin to confirm. The key action is ensuring your Product schema is complete (name, price, availability, brand, aggregateRating, shippingDetails, hasMerchantReturnPolicy), because AI agents read structured data before HTML.
What are Cloudflare's Redirects for AI Training?
A new Cloudflare feature (launched April 2026) that automatically converts your canonical tags into HTTP 301 redirects for verified AI training crawlers (GPTBot, ClaudeBot, Bytespider). Without it, training crawlers often ingest deprecated URLs like old product pages, causing outdated content to appear in AI responses. Available on all paid Cloudflare plans.
Is the old anthropic-ai user agent still valid?
No. Anthropic now uses three separate crawlers: ClaudeBot (training), Claude-SearchBot (search index), and Claude-User (live user queries). The old anthropic-ai directive does nothing — update your robots.txt to the current crawler names, especially allowing Claude-SearchBot for citations in Claude web search.
What does Perplexity's $500M ARR mean for e-commerce?
Perplexity now serves 100M+ monthly users and converts at 10.5% on e-commerce traffic (vs 1.76% Google organic). The company is pivoting from search to AI agents — meaning users increasingly ask an agent to research and buy on their behalf. Your product pages need to serve AI agents, not just human browsers: structured data, factual density, and proper schema.
Are ChatGPT ads replacing organic citations?
Not yet. OpenAI explicitly states ads are “separate and clearly labeled” and don’t influence organic answers. Paid tiers (Plus, Pro, Business, Enterprise) remain ad-free. But the geographic rollout (now US + AU + NZ + CA) signals ads are scaling. The practical takeaway: the free window for organic AI visibility is real but narrowing.
What's Claude Opus 4.7 and does it affect my AI visibility?
Claude Opus 4.7 (released April 2026) is Anthropic's flagship model with improved vision, instruction following, and agentic performance. For AI visibility, it means Claude’s web search and shopping agents are now more capable — they’ll pull from more sources and make more complex comparisons. Your site needs to be crawlable by Claude-SearchBot and Claude-User with complete structured data.