Key takeaways
- AI engines answer shopping questions with a short list of brands. If you're not named, you aren't considered for that question.
- They can only recommend what they can crawl, read and trust. Each of the three is something you can fix.
- Track three numbers together: Share of Answer (the outcome), AI Readability (what you control) and AI Visits (the traffic it brings).
- Fix crawler access, factual product data and structured data first. Then earn mentions on other sites.
On this page
What is AI visibility?
AI visibility is how often, and how favorably, AI engines such as ChatGPT, Claude, Perplexity, Gemini and Grok name your brand and products when people ask them for advice. It is judged on the answer itself, not on a ranking: either the assistant recommends you, or it recommends someone else.
The work of improving it goes by several names: answer engine optimization (AEO), generative engine optimization (GEO) or AI search optimization. They all describe the same goal: being the brand the answer names.
Why it matters for commerce brands
Shoppers now ask AI assistants the questions they used to type into a search bar: “What's the best linen duvet for hot sleepers?”, “Which running shoes are good for flat feet?”. Instead of ten links, they get a short answer that names a few brands and explains why.
- The answer is the shortlist. There is no second page to scroll to. A brand that isn't named is out of the running for that question.
- The reason is the pitch. The assistant explains its choice in a sentence or two. That sentence is your positioning, written by someone else.
- Not every recommendation becomes a click. Many shoppers take the name and look for the brand later. That's why being named matters, not just the visits AI sends you.
How AI engines choose which products to recommend
Engines differ in the details, but answers built with web search follow the same four stages.
They interpret the question
The engine turns the shopper's question into one or more searches (query fan-out), often adding details the shopper implied: budget, use, location, season.
They retrieve sources
It searches the web through its own crawler and search indexes. ChatGPT search, for example, uses OpenAI's OAI-SearchBot and third-party search providers. Google's AI results draw on Google's index. A page that can't be crawled or indexed can't be retrieved.
They read and extract facts
From the pages retrieved, the engine pulls specific facts: product names, prices, materials, sizes, policies, reviews. Facts stated plainly in the page's HTML and structured data are easier to quote than facts hidden in images, scripts or vague copy.
They weigh trust and write the answer
The engine combines what it found with what its model already knows from training. It tends to favor brands that several independent sources agree on: reviews, press, forums, marketplaces and comparison sites.
The three numbers to track
- Share of Answer
- What it tells youHow often AI names you for your buyers' questions, compared with competitors
- How it's measuredThe same set of buyer questions, asked to several AI engines on a regular schedule
- AI Readability
- What it tells youHow easily AI can read, understand and quote your pages
- How it's measuredPage checks for crawler access, structured data and factual product detail
- AI Visits
- What it tells youAI crawlers reading your site, and visitors sent by AI assistants
- How it's measuredCrawler user agents and AI referrers in your traffic
| Measure | What it tells you | How it's measured |
|---|---|---|
| Share of Answer | How often AI names you for your buyers' questions, compared with competitors | The same set of buyer questions, asked to several AI engines on a regular schedule |
| AI Readability | How easily AI can read, understand and quote your pages | Page checks for crawler access, structured data and factual product detail |
| AI Visits | AI crawlers reading your site, and visitors sent by AI assistants | Crawler user agents and AI referrers in your traffic |
Read them together. AI Readability is the part you control, Share of Answer is the outcome, and AI Visits shows whether it turns into traffic. A rising Readability score with a flat Share of Answer usually means the problem is off your site: other sources don't mention you yet.
What to fix first
Start with the fixes that unblock everything else, then work outward from your site.
Let AI crawlers in
Check robots.txt and your firewall or bot-protection rules. Search crawlers (OAI-SearchBot, PerplexityBot, Claude-SearchBot) and user-triggered fetchers (ChatGPT-User) put you in answers. Training crawlers (GPTBot, ClaudeBot, Google-Extended) feed future models. Blocking training is a business choice. Blocking search crawlers takes you out of answers, and CDN bot protection often does it without anyone noticing.
Put the facts in the HTML
Many AI crawlers don't run JavaScript. If prices, descriptions or reviews only load in the browser, those crawlers may never see them. Make sure the key product information is in the HTML your server sends.
Add structured data
Use Product markup with name, brand, price and availability, plus reviews when you have real ones. Add Organization markup for your brand, and FAQPage markup where a page answers real customer questions.
Write specific product copy
State materials, dimensions, fit, compatibility, care and who the product is for. AI can quote “100% European flax, 180 g/m², machine washable at 40 °C”. It can't do much with “premium quality you'll love”.
Make policies and contact details easy to find
Shipping, returns, warranty, contact information and secure checkout are trust signals. Publish them as plain pages that crawlers can reach.
Earn mentions on other sites
Reviews, press, comparison articles, forums and marketplace listings tell AI engines that others vouch for you. Keep your facts consistent everywhere: name, prices, what you sell.
Consider llms.txt, last
llms.txt is a proposed file that summarizes your site for language models. It's cheap to add, but no major engine has confirmed using it, so don't put it ahead of the steps above.
AI visibility vs. SEO
Good SEO is the foundation: AI engines rely on crawling, indexing, structured data and authority too. But ranking a page and being named in an answer are different goals.
- Goal
- SEORank a page
- AI visibilityBe named in the answer
- Unit of work
- SEOA keyword
- AI visibilityA buyer's question
- What the shopper sees
- SEOA list of links
- AI visibilityA short recommendation, with reasons
- Competitors
- SEOSites ranking for the keyword
- AI visibilityBrands named in the answer, including some you don't track
- Main measures
- SEORankings, clicks
- AI visibilityShare of Answer, AI Visits
| SEO | AI visibility | |
|---|---|---|
| Goal | Rank a page | Be named in the answer |
| Unit of work | A keyword | A buyer's question |
| What the shopper sees | A list of links | A short recommendation, with reasons |
| Competitors | Sites ranking for the keyword | Brands named in the answer, including some you don't track |
| Main measures | Rankings, clicks | Share of Answer, AI Visits |
How to measure progress
- Fix your question set. Use category questions that don't name you (“best linen duvet”) and comparison questions that name you and a competitor. Keep them the same from month to month.
- Ask more than one engine. Each one uses different sources, so a brand can lead in one and be missing from another.
- Look at trends, not single answers. AI answers vary from one run to the next. A share across many questions is reliable; one answer isn't.
- Log what you ship. Note the date of each fix so you can connect changes in the numbers to the work behind them.
- Watch the crawlers. A page has to be crawled before it can be cited, so AI crawler visits to the pages you fixed are the first sign that a fix is being picked up.