7 reasons AI doesn't recommend your products (and what to check today)

You ran the test from the previous article. You asked ChatGPT, Perplexity and Google AI Overviews for the three most important categories in your catalogue. Your store didn't show up, or it appeared once, unclearly.

Now you want the concrete reasons. There are seven. You can verify all of them today without touching the site code. If three or more apply, the problem is not fine-tuning. It is structural.

1. Product schema is incomplete or missing

AI assistants read your pages through schema markup. It is the standardized technical form in which you declare you are a store and your products are products. If it is missing or partial, the AI has to choose between guessing and skipping. It chooses skipping.

The required fields for a Product block that counts are: name, brand, sku, gtin (or gtin13, gtin14 as appropriate), image, description, offers with price, priceCurrency and availability. Missing any of brand, gtin or availability makes the product hard for the AI to cite, because it cannot confirm it is a real product, sold, identifiable.

You verify with Schema.org Validator (validator.schema.org). Paste the URL of a representative product page, run, read which fields are missing. If the tool does not detect a Product block at all, you have the foundational problem: your page is, for the AI, an article page, not a product page.

2. Organization or Brand schema is missing

Product schema makes you visible as a product. Organization schema (or Brand) makes you visible as an entity. The AI does not recommend brands it cannot confirm exist. Confirmation comes from Organization schema connected to your URL, with name, logo, links to verifiable social profiles, physical address if it exists, and ideally sameAs linking the brand to Wikipedia, LinkedIn, or another authority page.

The practical difference: two stores have the same products, the same prices, the same feed. Store A has complete Organization schema with a Wikipedia link. Store B does not. The AI recommends A, not because A is better, but because A is a confirmable entity.

You verify with the same Schema.org Validator on the homepage. Look for Organization block. Minimum fields: name, url, logo. Recommended: sameAs with social links, address, contactPoint.

3. Product names do not match how the buyer asks

In Google Shopping, a keyword-dense title works: "LED Bulb 12W E27 4000K Cool White". In AI Shopping, the assistant receives a natural-language question: "what white bulb should I buy for the living room". It needs to match the question to the product. A title like "LED 12W E27 4000K" does not tell the assistant this is a white bulb for the living room. It says a technical spec that assumes the buyer already knows the context.

The optimal title for AI contains: brand, product type written in natural language, main attribute visible to the buyer (colour, size, use), and optionally the technical spec for those who search for it. Example: "Philips Cool White LED Bulb for Living Room, 12W, E27 socket".

You verify manually: take 10 representative products from the catalogue. Read the title as a buyer who does not know the technical jargon. Do you understand what the product is and what it is for? If not, the AI does not understand either. The list of 10 titles you do not understand is the starting point of the rewrite.

4. Robots.txt or meta robots blocks AI crawlers

AI assistants have their own crawlers: GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, Google-Extended (Google AI), CCBot (Common Crawl, the training source for many models). If robots.txt blocks them (from historical caution, from copy-paste from another agency, from panic about "they will steal our content"), you are telling the AI exactly not to read you.

Stores that block AI crawlers with the intent of "protecting their content" sacrifice visibility in answers. The content is not protected. It is invisible. The competitor that does not block shows up in the ChatGPT answer. You do not.

You verify at https://[your-domain]/robots.txt. Look for Disallow rules targeting a user-agent like GPTBot, ClaudeBot, Google-Extended, PerplexityBot, CCBot, or * with Disallow on critical directories. If they exist, you have an active block. The decision to unblock is strategic, not just technical, but it must be made consciously, not left to inertia.

5. The site runs JavaScript-only and AI crawlers see empty pages

AI crawlers are less advanced than the Google crawler. Many do not execute JavaScript or execute it partially with a short timeout. If your page is built on a framework that renders everything client-side (many stores on Vue, React, or SaaS platforms without server-side rendering enabled), the AI crawler can receive almost-empty HTML and conclude the page has no content.

The quick diagnostic: open the product page, disable JavaScript in the browser (DevTools), reload. What do you see? If you see the product title, description, price, schema markup, you are fine. If you see a "Loading..." message or an empty page, you have a structural problem.

The fix is not simple (partial rewrite of the stack toward SSR or static generation), but the diagnostic is. The store that delivers complete HTML without JS is invisible to AI. The store that does not deliver does not exist in the AI's catalogue, no matter how good the products are.

6. Product data is not consistent between sitemap, schema and page

Your sitemap says product X has URL A, price B, last modified date C. The schema markup on the page says price D. The Google Merchant feed says availability "in stock", the page says "out of stock". The AI sees contradictions and loses confidence in the catalogue. The recommendation goes to the competitor.

The typical sources of inconsistency: an inventory system that updates the feed faster than the page (or vice versa), displayed price that differs from the schema markup price because they are generated from different CMS fields, "false-true" availability on a product's variants (size, colour) when only one is in stock, sitemap modification date that does not correspond to a real change.

You verify with a cross-reference test: pick 5 products, open the page, verify the displayed price, verify the schema markup in DevTools (View Source, search for application/ld+json), verify the sitemap (/sitemap.xml) for the last modification date, verify the Google Merchant feed for the same products. Three sources, the same data. If a difference appears on any of the five, the problem is systematic, not isolated.

7. Your brand has no confirmable authority

Schema is the form. Authority is the content. A store with perfect schema but zero external mentions (Wikipedia, press articles, reviews on known platforms, profiles on industry sites) is not recommended. The AI has no way to confirm you are a real store, with history, with reputation.

Building authority is not technical. It takes time and real mentions. The fragments that count most: a Wikipedia page (even short, but it has to be approved per Wikipedia's rules), signed press articles that cite you as a source or example, reviews on Trustpilot/Yelp/Google Business with sufficient volume to be credible, presence in "top X in [category]" lists on industry sites.

You verify simply: open Google and search the exact brand name. Do you see a Wikipedia page, press articles from the last 12 months, an active Google Business Profile with reviews? If two of these three are missing, you have an authority problem that schema markup does not solve. It is solved with PR and citable content.

How to check all 7 in one sitting

Three free tools and one free hour cover the 7 checks:

• Schema.org Validator (validator.schema.org) for reasons 1, 2, and partly 6.
• Browser DevTools (F12 or Cmd+Option+I) with JavaScript disabled for reason 5, plus View Source for the cross-reference at reason 6.
• Manual Google search for the brand name and for robots.txt rules, covering reasons 4 and 7.

The list you have at the end is your map. The 7 reasons are not fixed in parallel. They have an economic order: reasons 4 and 5 are total blockers (without them, the rest does not matter). Reasons 1, 2, 6 are technical foundation (without them, authority does not transmit). Reasons 3 and 7 are amplifiers (with the foundation in place, they decide how often you appear, not whether you appear).

Schema is the form. Authority is the content. Neither works alone. The stores that consistently appear in AI answers have both.

What the AI-Ready Catalog Audit does

The 7 checks above tell you where you stand. They do not tell you the order of fixes, the cost of each fix, what to handle internally and what to outsource. That is what the AI-Ready Catalog Audit is for.

In 5 working days, we map for your account: 5 categories manually tested on ChatGPT, Perplexity and Google AI Overviews; schema markup verified on 20 representative pages; Merchant feed structure compared against schema and page; brand authority evaluated against the competitors that appear in answers for your category; a fix plan in order of economic impact.

If 3 or more of the 7 reasons above apply to your store, the audit gives you in 5 days what you would obtain in 3 months of fragmented internal work: a clear order and an estimate of the gain per fix.

What we do next

In the next article in this series, we shift perspective. From "how do I show up in AI answers", to "what happens when the buyer no longer asks the question and the AI agent buys in their place". Agentic commerce becomes mainstream in 2026 and changes the economics of customer acquisition. Preparing for AI Shopping is preparing for the world in which AI agents are buyers, not intermediaries.