AI SEO (GEO): a 12-point checklist for online stores
Updated 2026-08-03
What AI SEO is, and why it matters now
AI SEO — often called GEO, Generative Engine Optimization — is getting your store ready for a world where products are recommended not by a results page but by an AI assistant: ChatGPT, Perplexity, Gemini or Claude. The goal is the same as classic SEO — be in the answer — but the rules differ: the assistant needs a machine-readable price, clean structured data and open access for bots.
How real is this today? In our AI traffic index (13 Jul – 1 Aug 2026, three panel stores), one AI vendor alone sent 606,017 requests to a single store — roughly 34,000 a day — and bots read 73,715 product pages at the panel's most machine-readable store. Yet our rating of 63 MD/RO stores shows a median score of just 38 out of 100. The bot demand is already here; most stores aren't ready for it. That gap is your opportunity.
How AI readiness is scored
Botmetria's free URL audit returns a 0–100 score from five weighted categories:
| Category | Weight | What it checks |
|---|---|---|
| Discovery | 20% | robots.txt, sitemap.xml, llms.txt, product feed, /.well-known/ucp |
| Product data | 30% | JSON-LD Product (price, currency), availability, GTIN/SKU, product images open to bots |
| Logistics | 15% | shipping and returns readable by AI |
| Hygiene | 15% | how the home page answers GPTBot, response time, redirects |
| Agent Markdown | 20% | clean text/markdown for agents |
The bands are simple: ≤40 is red, 41–70 yellow, above 70 green. The checklist below covers all five categories, plus three items the score doesn't measure directly but that turn bot traffic into orders.
The 12-point checklist
- robots.txt is open to the AI bots you want. Allow
GPTBot,OAI-SearchBot,PerplexityBot,ClaudeBotand the rest — otherwise the assistant physically can't read your products. Never blockGooglebotorBingbot, though: that kills classic SEO. Full walkthrough: robots.txt and GPTBot. - llms.txt sits at the site root. It's a short Markdown map of your store — what you sell, where the key pages are, your shipping and return rules — so the model gets the gist without wading through HTML. See llms.txt for stores.
- sitemap.xml is current and complete. Without a map, bots find your products slowly and only in part. Check whether crawlers can see you with the free audit.
- Content is visible without JavaScript. Crawlers that don't run JS see an empty page instead of your price and description. Serve the key data in HTML. Here's why: ChatGPT can't see your site.
- JSON-LD Product with price, currency and availability. It's from
schema.org/Productwithoffers{price, priceCurrency, availability}that an assistant quotes your price and "in stock" status. See JSON-LD Product and availability. - Fresh prices and a product feed. If the feed lags, AI will quote an old price and the customer arrives with the wrong expectation. Keep it in sync. See product feed and why ChatGPT shows old prices.
- Shipping and returns are marked up. Delivery times, cost and return terms must be machine-readable, not buried in a banner. How to mark them up: shipping and returns.
- Product images don't return 403/406. Bots request a photo and get denied, so the product stays "faceless" — and the audit counts it against your Product data score. How to fix it: images returning 403/406.
- A markdown twin of your pages. Clean
text/markdownfor agents reads faster and more accurately than heavy HTML. What it is: Markdown for agents. - External mentions. Models recommend what others write about — reviews, catalogs, forums. How to build them: how to get into ChatGPT.
- AI-channel attribution is set up. Orders from ChatGPT land in Google Analytics as "direct" — the referrer is stripped and the channel is undercounted. Botmetria tags AI referrals on arrival; the WooCommerce plugin ties orders to their source automatically, and with the pixel you add one call on the order confirmation page. Details: ChatGPT orders as direct in GA and AI-assistant referrals.
- Audit on a regular cadence. AI readiness isn't a one-off — prices, feeds and markup change. How to measure it systematically: the AI-readiness methodology.
Where to start
If you're short on time, follow the category weights. Start with Product data (30%) — JSON-LD with price and availability — then the images item; together they give the fastest score gain. Next close Discovery (robots, llms.txt, sitemap) and Markdown, another 40% of the weight. Leave attribution and external mentions for later, but don't skip them: without them you won't see the payoff.
Common mistakes
- Blocking every bot "just in case." You also shut out the ones that bring buyers. Verify bots by their published IP ranges, not by name, and block only the fakes — how to verify GPTBot.
- Expecting orders to show up as "AI" in GA. They won't — the channel collapses into "direct." Set up attribution before the first sales, not after.
- Marking up one product and forgetting it. The score lives with your catalog: a new batch without stock or broken images drags it right back down.
FAQ
How is AI SEO (GEO) different from regular SEO?
Classic SEO competes for a spot in Google's results; AI SEO competes to have the assistant name your product, price and availability right in its answer. That needs machine-readable structured data and bot access, not just keywords and links.
Where should a small store start?
With Product data — JSON-LD carrying price, currency and availability (30% of the weight). Missing or incomplete product markup is the most common failure in our rating, and fixing it gives the fastest score gain; product images that return 403/406 to bots are checked in the same category.
How do I know which items I've already covered?
Run Botmetria's free URL audit — it returns a 0–100 score across the five categories and flags specific problems for the items it measures (robots.txt, llms.txt, sitemap, product markup and images, shipping and returns pages, bot access, markdown). External mentions and attribution aren't part of the score.
Will limiting AI bots hurt my regular SEO?
No, as long as you leave Googlebot and Bingbot alone — those must never be blocked. Verify bots by their published IP ranges and block only the fakes.
Related guides
Check your store for free
A 60-second audit shows how ready your store is for AI agents — and exactly what to fix first.