Ecommerce SEO Audit: What to Check for Google and AI

An ecommerce SEO audit used to mean one thing: crawl the site, fix what the crawler flags, check the product pages. That is still a sound start. It just no longer covers the places where most online stores are now found.

Shopping ads, free listings, Performance Max and AI assistants answering shopping questions do not read your page copy first. They read the product feed behind it. So an audit that stops at the site can report a clean bill of health while half your catalog sits unseen. This guide walks through the four layers worth checking, in the order that wastes the least time, and ends with a way to rank what you find.

A note on what this is. A checklist tells you what to do. An audit tells you what is wrong. If you want the “do this” version for product pages, Product Page SEO in 2027 covers it. This post is the diagnosis that comes before it.

How to Run an Ecommerce SEO Audit in Four Layers

Work from the outside in: can machines reach the store, is each page sound, does the product record say the right things, and can an AI system use it. Each layer depends on the one before it. A beautiful feed means nothing if the pages it points to are blocked, and a perfect page means little if the record behind it is thin.

Before you start, open a spreadsheet with four columns: URL or product, what you found, which layer it belongs to, and severity. Every check below ends in a row in that sheet. If a check passes, you move on without writing anything down.

For pages, sample rather than inspect everything. Pick 20 to 30 products spread across best sellers, mid-tier items and the long tail. The long tail matters most, because that is where templates break and nobody looks.

Layer 1: Site Health, Crawling and Indexing

This layer asks whether Google can reach and understand the store at all. Open Google Search Console and check these in order.

• Page indexing report. Look at the pages marked “Crawled, currently not indexed” and “Discovered, currently not indexed.” On a store these are often product pages with thin or near-duplicate content, or filter URLs that should never have been crawled.
• Product snippets and Merchant listings reports. Search Console shows structured data errors for product pages here. Missing price or availability in your markup is the usual culprit.
• Core Web Vitals. Google’s thresholds for a “good” result are LCP under 2.5 seconds, INP under 200 milliseconds and CLS under 0.1. Check mobile first, since that is where most shoppers arrive.
• Duplicate and faceted URLs. Color, size and sort parameters can multiply one product into dozens of crawlable URLs. Confirm each product has one canonical URL and that variants point to it deliberately.
• Sitemap and robots.txt. Make sure the sitemap lists live product URLs only, with no redirects, no out-of-stock dead ends you meant to remove, and nothing blocked that you want indexed.

Findings here are usually Blockers or Leaks. A product that cannot be indexed cannot rank, and no amount of copywriting changes that.

Layer 2: Page Quality on the Product Pages

Now look at the sampled pages the way a shopper and a search engine would.

Start with the basics: one H1 that names the product, a title tag that matches the product the same way the H1 does, and a description written for the product rather than pasted from the manufacturer. Manufacturer copy appears on every competing store, which is exactly why it ranks nowhere.

Then check what a machine can lift out of the page. Dimensions, materials, compatibility and care instructions should be stated as plain facts, not buried in adjectives. Reviews, if you have them, should be marked up and visible. Images should load fast, carry descriptive alt text, and be free of watermarks or promotional overlays, which Google’s feed rules also prohibit.

Check internal linking too. Every product should be reachable from a category page within a few clicks, and category pages should link back to their best performers. Orphan products rarely get indexed or rank.

This is the layer traditional audits cover well, and most store owners stop here. The next layer is why that is a mistake.

Layer 3: The Product Feed Audit Most Audits Skip

Your feed is the record Google Merchant Center holds for each product, and it is what Shopping ads, free listings and Performance Max actually read. Page and feed are two separate documents describing one product, and audit findings live in the gap between them.

Open the Diagnostics page in Merchant Center first. It lists account-level issues, feed-level problems and item-level disapprovals. Anything disapproved is a Blocker by definition: those products are not serving.

Then audit the things Diagnostics does not rank for you:

1. Price and availability parity. Compare feed values against live pages for the sampled products. Google treats mismatches as a data quality violation and can disapprove items over them. Sale periods are when this breaks, because pages change faster than feeds refresh.
2. Title structure. Google allows up to 150 characters, but the visible portion is shorter, so the words that matter should come first: brand, product type, the attributes buyers filter on. A title like “Blue Shirt Men” is a Leak on every surface at once.
3. Attribute completeness. Count the products missing color, size, material, gender, age group and product category. Missing attributes limit which searches a product is eligible for, and our guide to how Google punishes missing feed attributes maps exactly how.
4. Identifiers. List products without a GTIN or MPN. Do not fill the gaps with guesses: a made-up GTIN is a policy violation. Source the real code from the manufacturer or GS1. Google’s own figures put correct GTINs at roughly 20% more clicks.
5. Description depth. Thin feed descriptions starve every downstream surface. If the description is one line, the AI system reading it has one line to work with.

An audit of the page tells you what Google can crawl. An audit of the feed tells you what Google, Performance Max and AI assistants can actually use.

Run the parity and attribute checks across the whole catalog, not just the sample. Feed problems are systematic: one broken template or one missed mapping repeats on every product it touches, so a sample understates the damage.

Layer 4: AI Readiness

The last layer asks whether AI shopping surfaces can find, read and trust the product. Be realistic about what you can verify. No audit can confirm that an assistant will recommend a product, but it can confirm that nothing is in the way.

Check robots.txt for rules that block the crawlers these systems use. OpenAI documents OAI-SearchBot for search results and GPTBot for model training, and Perplexity documents PerplexityBot. Blocking a search crawler can keep your pages out of that product’s answers. Whether to block training crawlers is a business decision, but know which one you are making. Google has said its AI features in Search follow normal Search eligibility, so Googlebot access and a healthy index come first there.

Then test the record itself. Does each product state its facts as structured data, or only in prose? Are the conversational fields filled in, the highlights, details and common buyer questions Google added for AI surfaces? Google says these do not affect product approval, so they never show up as errors. They only show up as a gap in how much an assistant has to work with. The reasoning is laid out in our piece on why AI assistants ignore most product feeds.

Finally, run a few real prompts. Ask the assistants for the kind of product you sell, in the way a shopper would phrase it, and note who shows up and why. This is anecdotal, not measurement, but it tells you which competitors are structured better than you.

Score and Rank What You Found

A finished audit can produce hundreds of rows. Without a ranking rule, you will fix the easy ones and feel productive while the Blockers sit untouched. Sort every finding into one of three bins.

• Blocker. The product cannot serve or be indexed: disapprovals, blocked crawlers, noindex mistakes, broken canonicals. Fix these first, today if possible.
• Leak. The product serves, but loses to better-described competitors: weak titles, missing attributes, price mismatches, thin descriptions. Fix these in order of product revenue.
• Polish. Real improvements that move nothing alone: alt text refinements, extra internal links, small copy edits. Batch these last.

For the feed layer, a per-product score does the sorting for you. A UCP Score rates each product from 0 to 100 on how complete and AI-ready its record is, so you can sort the catalog from weakest to strongest instead of reading every row. The worked example in From UCP Score 28 to 92 shows what that triage looks like on a real catalog.

Finally, set a re-audit date. Quarterly for the full run, monthly for the feed layer. Catalogs drift, plugins update, and a sale that changes 300 prices overnight can undo a clean audit in a week.

Conclusion

A good ecommerce SEO audit is not longer than the old kind, just wider. Site health and page quality still decide whether you can be found in web search. The feed and AI layers decide whether you appear everywhere else, and they are the layers most store owners have never looked at.

Run the four layers in order, log every finding, rank them as Blocker, Leak or Polish, and fix from the top. UCP Radar handles the feed layer for you: it checks your catalog against 50+ Google Merchant Center rules, gives every product a UCP Score, shows which attributes and identifiers are missing so you can source the real ones, and builds optimized supplemental feeds from the result. If you want the feed half of your audit without the manual spreadsheet work, start with a free trial and see which products need attention first.