Why Your Reviews Are Missing From Google AI Overviews

How does an AI Overview decide what to pull in?

An AI Overview is not a fresh crawl of the web. It is a Gemini layer that summarises what Google has already indexed, plus a handful of live retrievals to confirm or extend the answer. That single fact decides most of what follows: if a passage was never in the index, it was never a candidate for the Overview, no matter how good the review is.

So the foundations carry over from classic SEO. The page has to be crawlable, the review text has to be in the HTML Google stored, and the question your buyer typed has to map cleanly onto a passage on your page. The Overview then lifts the part that answers most directly and attributes it to a source it can name.

Why are my reviews missing when my page ranks fine?

Ranking and extraction are two different tests, and most stores pass the first and fail the second. Your product page can rank for its keyword on the strength of its title, copy, and links, while the reviews themselves never enter the index because they load after the page does.

Most review apps inject reviews through a JavaScript widget. A shopper sees stars and quotes. The stored HTML that feeds the Overview often holds an empty container where that social proof should be. The page ranks; the reviews are simply not there to quote.

What actually gets quoted in an Overview?

Overviews favour the passage that answers the prompt most directly, and they favour it early. A front-loaded sentence that states the answer in the first lines of a section is pulled far more readily than the same point buried three paragraphs down.

For reviews, that means specificity wins. "Great quality, love it" answers nothing a buyer asked. "These ran half a size small and the arch support held up over a marathon" answers a real question, and is the kind of sentence the Overview will lift and attribute.

• Passages placed in the first portion of a page or section, not the footer.
• Review text that names a use case, a condition, or a measurable outcome.
• Content that maps onto a question phrasing rather than a star rating alone.
• Sources Google can attribute cleanly, with a product the rating clearly belongs to.

Does structured data still matter if Gemini reads the page?

Yes, and arguably more than for plain blue links. Star-rating structured data does not just earn the rich result; it tells the Overview which product a rating belongs to, so the system can attribute "rated 4.6 by 2,300 buyers" to the right item with confidence rather than guessing from nearby text.

The extraction layer rewards being unambiguous. Schema removes ambiguity. Without it, the model has to infer the link between a rating and a product from page layout, which is exactly the kind of inference it will skip in favour of a competitor who made it explicit.

• Mark up product reviews with Product and AggregateRating schema.
• Keep the rating, count, and product name consistent between schema and visible HTML.
• Do not mark up reviews that are not visible on the page; Google treats that as a violation.

How is this different from getting cited by ChatGPT or Perplexity?

The classic SEO foundations carry more weight here than in any other answer engine, because the Overview is built directly on the index you already optimise for. Crawlable HTML and structured data are not nice-to-haves; they are the gate.

That is the trade-off. Tools like Perplexity lean harder on third-party corroboration and live retrieval, so a brand with strong external profiles can surface even with a weak own-site setup. An Overview is less forgiving of an unreadable page, but more rewarding of one you have already made extractable. Get the on-page work right and you are a candidate for both.

How long until a fix shows up in Overviews?

It tracks the index, so it is slower than an instant test but more durable than a live-retrieval citation. Once your reviews are in the server HTML and marked up, the change has to be recrawled and reindexed before an Overview can reflect it, which is weeks rather than days for most stores.

The upside is stability. Because the Overview reads from the index rather than a volatile live fetch, an inclusion you earn tends to hold while the underlying page does, instead of decaying on its own clock the way a pure AI citation can.

What this adds up to

For Overviews the order is clear: readable first, marked up second, specific third. Readable means reviews in the server HTML so they enter the index at all. Marked up means star-rating structured data so the Overview can attribute the rating to the right product. Specific means review text that answers the question a buyer typed.

Most review apps were built for the on-page shopper and stop there, which is the exact gap BetterReviews is built to close: getting the reviews you already have readable, corroborated, and quoted by the engines buyers now ask.