Your store has two audiences: the people who buy and the robots that decide where you show up. And the second one is no longer a minority: according to Cloudflare's July 2026 bot report, more than half of web traffic is no longer human. Among those robots are the crawlers that feed ChatGPT, Copilot and Perplexity — and here's the key part: if those bots can't read your store, for AI assistants your store doesn't exist. Working on that visibility has a name: GEO, Generative Engine Optimization — structuring your content so AI systems find it, understand it and cite it in their answers.
The good news: you don't need to redesign anything. You need to review two layers of accessibility, and the trickiest one can be audited from Cloudflare in 15 minutes.
Human layer: a person can read your store. Alt text on product photos, a clear heading hierarchy (one H1, an H2 per section), clean HTML and real text — no prices or measurements buried inside an image. This is classic accessibility, the screen-reader kind. And it turns out an LLM reads your store just like a screen reader does: it doesn't see the photo, it reads the alt text; it doesn't see the design, it reads the heading structure. What you wrote for people is exactly what the model needs to understand your catalog.
Crawler layer: the robot can get in. Perfect HTML is useless if the bot bounces before entering. This layer asks for simple things: don't block it, don't ask it to solve a captcha (bots don't solve captchas — they leave), don't hide the catalog behind a login, and serve content rendered from the server: GEO guides agree on avoiding client-side rendering for key content, because many crawlers don't execute JavaScript.
The best part: every improvement in the human layer improves the crawler layer for free. They're not two projects: they're the same one.
Roughly one in five websites goes through Cloudflare, so its configuration acts as a gatekeeper: it decides what Googlebot, Bingbot and AI bots get to see. These five signs suggest your gatekeeper may be turning away the wrong guests:
Disallow someone added years ago "for the scrapers" may now be cutting off the AI search bots. If you do not know where you stand, the bot accessibility test tells you what happens layer by layer.And one myth to dismantle: blocking Google-Extended doesn't remove you from Google's AI Overviews. According to searchVIU, Overviews are fed by Googlebot, the same crawler as always; Google-Extended only controls whether your content is used to train Gemini.
You don't need to be technical: it's four screens.
search=yes, ai-train=no: "search me and cite me, but don't train on my content for free". Careful: these are declared preferences, not a technical wall. What each signal actually declares, and why no crawler honours it yet, we take apart in Content Signals in robots.txt.Want the fine detail on WAF, cache and verified bots? It's in the Cloudflare technical guide for SEO and GEO. Your domain doesn't go through Cloudflare yet? Start with the step-by-step implementation guide.
The crawler getting in is half the game. The other half is that the index where the AI searches has your catalog up to date — today's prices, today's stock. That rests on two pieces: a healthy sitemap, the full inventory search engines re-read every so often, and the instant notice of every change via IndexNow, the push protocol that goes straight to Bing — the index that feeds ChatGPT and Copilot.
Cloudflare's own Crawler Hints uses IndexNow, but it fires on heuristics: it infers from the cache that something changed. IndexNow Connect works with the exact signal: it listens to every addition, change and removal in your catalog in real time and sends the precise URL to IndexNow the moment it happens — without touching your store's code. Readability (Cloudflare) plus freshness (IndexNow) is the complete combination.
If you checked everything, your store is readable and fresh. If you're missing the last two, that part can be automated.
Yes, and GEO too. Alt text, heading hierarchy and real text are the same signals screen readers, classic search engines and LLMs use to understand a page. It's a single investment that pays off with all three audiences.
Check in Cloudflare's AI Crawl Control whether OAI-SearchBot is allowed and logging activity, and in Security → Events that it's not getting challenges or blocks. If your zone was created after July 2025, assume it's blocked until proven otherwise.
No: your prices are already public to any visitor, bots or not. What you do achieve by blocking everything is disappearing from ChatGPT and Perplexity answers — the GEO paradox. The smart move is discriminating by purpose: ai-train=no for training, and allow the search bots.
For visibility, no: neither search engines nor AI assistants cite what they can't read, so a login-gated catalog is invisible in ChatGPT, Copilot and Perplexity. If you need private prices (wholesale, for example), keep at least the product pages public with the general information.
Connect your store for free → — IndexNow Connect notifies Bing of every catalog change the moment it happens, without touching your store's code.