AI Search exposes every shortcut you took for Google

A site ranks fine in Google. Decent traffic, decent positions. Then someone runs a few brand prompts in ChatGPT or Claude and gets nothing. Or worse, gets a clean citation for the competitor down the road.

The site owner is confused. We're #2 in Google for that query. How are we invisible here?

The answer is almost always the same. The site has been getting away with shortcuts for years that Google quietly forgave. AI Search doesn't.

Thin product pages. Duplicate meta descriptions across whole categories. Prices hidden behind a "request a quote" form (mostly on B2B sites). No real entity definition anywhere in the markup. None of it completely killed Google rankings, so nobody fixed it.

Now Claude won't cite the page. ChatGPT won't recommend the product. Google's own AI Overviews skip it for a competitor with cleaner data.

Google was patient. AI is not

Google built its index over two decades. It learned to handle messy sites. It can stitch together thin content with internal links and external signals, guess what a duplicate-described page is about. It tolerates a lot.

AI engines don't operate this way. They retrieve a small set of pages, extract structured information, and synthesize an answer in seconds. There's no time for charitable interpretation.

If the page doesn't have a clear answer, they pick a different page. If the price isn't in the markup, they pick a competitor. If the schema is missing, they fall back to text parsing – always less reliable than reading a clean JSON-LD block.

Pages with valid schema markup show up more often in AI Overviews. The pages that win citations are the ones that make the engine's job easy.

Thin content and duplicate descriptions

Programmatic e-commerce is the worst offender. 10,000 product pages, the same 80-word description mildly templated, the same meta description across half the products with one variable token swapped in.

Google was lenient. AI engines aren't. Duplicate content forces them into a binary choice about which version is authoritative – and if they can't decide, they often skip all of them. Two thin pages targeting the same query don't compete for AI citations. They cancel each other out.

The fix is the one SEOs have been recommending for a decade. Consolidate the thin pages into one strong page, or make them actually unique. There's no third option that works for AI.

If you can't write unique copy for every product, lean on structured data. A page with sparse text but rich Product schema – price, availability, reviews, specs – is far more useful to an AI engine than the same page with the schema missing. Schema is how you turn "thin" into "machine-readable enough."

Weak entity signals are the silent killer

Ask ChatGPT about a competitor and you get a clean two-paragraph description, list of products, target audience. Ask about your own brand and you get "I don't have detailed information about [your company]."

The cause is almost always the same. No Organization schema with sameAs linking to social media profiles and other sources. No consistent entity presence anywhere beyond the site itself. The brand exists as a website. It does not exist as a thing in the public knowledge graph.

For Google rankings, this didn't matter much. For AI citations, it's the gating factor. AI engines need to confidently identify what entity you are before they recommend you. If signals are inconsistent or absent, the engine hedges. Or skips you.

The price you didn't put on the page

A product page with no price. A category that lists products with "Contact us for pricing" instead of a number. Prices loaded by JavaScript after the page renders.

For Google, this was annoying but not fatal. The page ranked. Forms got filled. Sales did the rest.

For AI Search, the missing price is often the entire reason the page isn't cited. ChatGPT Shopping, Perplexity's product modules, AI Overviews for commercial queries – they all need a price to do their job. The query is "best [product] under [budget]." Without the number, you're not on the list.

I've audited B2B ecommerce websites where the locales that show prices outperform the locales that hide them by a wide margin on commercial queries. Same products. Same content. The only difference is the number on the page.

The pattern across all four

None of these are new problems. Unique titles and descriptions, real prices on product pages, substantive content, clean schema, entity disambiguation. All of it was on the SEO best-practice list in 2019.

What's new is the cost of ignoring them. Google made SEO forgiving. Too forgiving. A whole generation of sites learned you could skip the boring fundamentals and still grow if your authority was strong enough.

AI Search broke that bargain. The engines pick the page with the clearest answer, the cleanest structured data, and the actual numbers. Sometimes the page that wins the citation isn't even on page one of Google.

The sites doing well in AI Search were doing the right thing for Google five years ago.