How to audit an ecommerce site for AI search visibility

More and more shopping journeys now start inside an AI answer rather than a list of blue links. Someone asks ChatGPT for the best waterproof walking boots under a hundred pounds, or asks Google’s AI to compare two coffee machines, and the AI names specific products and brands. If yours is not among them, you are losing sales at the very top of the funnel and, in most cases, you have no idea it is happening. AI search visits grew dramatically over the past year while traditional search barely moved, and the visitors arriving this way tend to convert well, which makes this a channel worth measuring rather than guessing at.

An AI search visibility audit tells you where you actually stand: whether AI systems can reach your pages, whether your content is built to be quoted, and whether you are being recommended, ignored or losing out to competitors. Here is how to run one for an ecommerce site, step by step.

First, understand what you are auditing for

A traditional SEO audit asks whether your pages can be crawled, indexed and ranked. An AI search audit asks a different question: can generative engines retrieve your content, understand it, trust it and use it inside an answer? The goal has shifted from ranking in a list to being included in the response.

That means an AI visibility audit looks across four connected layers. Can the AI systems physically access your content? Is that content structured so it can be extracted and quoted? Does your brand carry enough authority across the web to be trusted as a source? And are you measuring any of it in a way that lets you improve? A good audit works through all four, so we will take them in turn, with the specific quirks that ecommerce sites bring to each.

Layer one: Can AI systems actually reach your content?

None of the rest matters if the AI crawlers cannot get to your pages, and this is where ecommerce sites trip up most often.

The first thing to check is your robots.txt file, because AI systems use their own crawlers, with names like GPTBot, OAI-SearchBot, PerplexityBot and Google-Extended. If your robots.txt blocks these, whether deliberately or through a rule someone added without realising, you have made yourself invisible to those engines before the audit even begins. Confirm which bots you are allowing and make a conscious decision about each, rather than leaving it to a default nobody chose.

The second, and the biggest technical issue specific to ecommerce, is JavaScript rendering. Many modern storefronts, particularly on platforms and frameworks that build the page in the browser, load their key content through JavaScript. Some AI crawlers do not execute JavaScript at all, which means they may arrive at a product page and see almost nothing: no price, no description, no specifications. To test this, view a product page with JavaScript disabled, or fetch it as raw HTML, and check whether your core content is actually present in the source. If your product details only appear once scripts run, a good deal of AI search cannot see them.

While you are here, confirm the fundamentals that also feed AI access: a clean XML sitemap, no accidental NOINDEX tags on important pages, and reasonable page speed, since slow, heavy pages hamper crawling of any kind.

Layer two: Is your content built to be extracted?

Once systems can reach your pages, the question becomes whether your content is in a form they can lift a clear answer from. This is where most ecommerce sites have the furthest to go.

Product and collection page content

AI systems struggle to say anything useful about a product page that is a bare image, a price and a manufacturer’s boilerplate description. They favour content that answers real questions. So audit your key product pages for original descriptions that cover the things buyers ask about: materials, dimensions, use cases, care, what makes the product suitable for a particular need. Manufacturer descriptions copied across dozens of retailers give an AI no reason to cite you specifically over anyone else selling the same item.

Collection pages deserve the same scrutiny and rarely get it. A collection page with a paragraph of useful, original context is far more likely to be drawn on for a category-level question like “best budget road bikes” than a bare grid of products.

Structure and clarity

Because AI systems retrieve and quote passages rather than whole pages, check that your content is broken into clear, self-contained sections with descriptive headings, each answering a specific question. Content that leads with a direct answer and then expands is far easier to extract than a dense, meandering block. This is the semantic chunking principle applied to a store, and it is worth reading our fuller piece on semantic chunking alongside this audit.

Structured data

Schema is close to essential for ecommerce AI visibility, because it hands the machine the facts in a form it does not have to interpret. Audit your product schema to confirm it includes price, availability, brand, and review ratings, and that it accurately matches what is on the page. Check for organisation and breadcrumb markup too. Incomplete or missing structured data is one of the most common issues that limits a brand’s AI visibility, and our guides on schema markup and rich results cover how to get it right.

The questions your customers ask

AI search is heavily driven by real, conversational questions. Audit whether your site actually answers them: buying guides, comparison content, and useful FAQ material around your products. A store that only has product and collection pages, with no content addressing how to choose between options or how to use a product, has little for a question-led engine to work with.

Layer three: Does your brand carry authority off-site?

This is the layer that catches people by surprise, and it is one of the most important. AI systems do not judge your brand on your own website alone. They weigh your entire presence across the web when deciding whether to trust and cite you, and coverage across multiple platforms materially increases your chances of being cited.

So part of the audit happens away from your site. Check whether your brand and your key products are mentioned on the third-party sources these engines lean on: reputable review sites, industry publications, buying guides, comparison sites and, where relevant and warranted, the major reference sites. Look at your reviews, both their volume and their sentiment, since these strongly influence whether an AI recommends you. And look honestly at your competitors: run the buying questions that matter in your category and note who gets named. If AI consistently recommends the same rivals, their off-site authority is usually a big part of why. Those gaps become your off-site roadmap.

Layer four: Measure it, then keep measuring

You cannot improve what you are not tracking, and AI visibility is not a one-off check because these engines change constantly.

The practical method is prompt testing. Build a list of the questions a real customer might ask that should surface your products, the category comparisons, the “best X for Y” queries, the specific problems your products solve. Then run them across the main engines, Google’s AI Overviews and AI Mode, ChatGPT and Perplexity, and record what happens: whether you are mentioned, which of your URLs are cited, how you are described, and which competitors appear alongside or instead of you. Checking a single engine is close to meaningless, because each has different sources and behaviour, so test across several.

Log all of it so you have a baseline to beat, then repeat on a regular cadence, quarterly as a fuller audit with lighter checks in between. Google’s Search Console now also surfaces dedicated AI performance data, which we covered in our piece on the new AI metrics in Search Console, and that gives you a useful, first-party view of your AI impressions to sit alongside the manual prompt testing.

Turning the audit into action

An audit is only worth running if it changes what you do next. Once you have worked through the four layers, you will typically have a clear, prioritised list: fix any crawler access problems first, because they block everything else, then address the content and schema gaps on your highest-value product and collection pages, then build the off-site authority and content that earns trust over the longer term. Measured, prioritised and repeated, this is what moves a store from invisible to cited in the answers that increasingly decide where people shop.

If you would like us to run an AI search visibility audit on your store, or build the fixes into a wider SEO services and technical SEO strategy, get in touch with the team at TAL and we will show you exactly where you stand and what to fix first.

All Articles