Review management. Make your reviews readable, recent and real.

Authority · Review management

When an assistant recommends a brand, it checks what customers say. We make sure it can find your reviews, read them and trust where they come from.

What is review management for AI search?

Review management for AI search is making sure the reviews about your brand are collected on independent platforms, readable by machines and answered by you. AI assistants use reviews as the second opinion in an answer: the score, the number of reviews, their age and what customers actually wrote.

Reviews are the source next to your name

Trustpilot and SEER Interactive analysed more than 800,000 AI responses. Review and trust sites were the second most-cited source category. 59% of AI recommendations included a score or star rating as evidence. Trustpilot sells review software, so read that study as a vendor's view. Our own sampling points the same way.

  • On one retail client's brand prompts, a review platform was cited in 54% of answers. The brand's own site was cited in 46%.
  • In one audit, ChatGPT discounted a 9.6 score in its answer. The reason it gave: the score came from a single platform where the company invites its own customers.
  • In a home-improvement panel, a review platform is the third most-cited domain in ChatGPT and the second in Gemini. It is cited more often than any company site in the category.

A high score on one platform is fragile. A spread of recent reviews on independent platforms holds up.

Fix in this order

1. Operational. Read what the reviews complain about and fix it. Delivery, service, billing. No reply strategy survives an unsolved problem, and the model quotes the complaint either way.

2. Editorial. Make your reviews and your facts readable for machines. This is where most brands lose reviews they already have.

3. Reputational. Reply to every review. Ask every customer, on more than one independent platform. Keep the newest review recent.

Can a machine read your reviews?

Often it cannot. Three checks from our audits:

  • Reviews loaded by a widget. Most AI crawlers do not run JavaScript. If your review page loads its reviews through a script, the crawler sees an empty page. We audited a brand that reports over 100,000 reviews. None of the review text was in the HTML.
  • A score without evidence. "Rated 8.6" with no count, no dates and no review text gives a model nothing to verify.
  • Markup in the wrong place. Google does not show review stars for reviews a business publishes about itself on its own organisation or local business pages. Product pages are different.

What we build instead:

  • Review pages rendered on the server, with real review text, dates and a written summary of themes.
  • Product markup with aggregateRating and Review on product pages, where the rules allow it.
  • A facts page that states policies, prices and guarantees plainly, so the engine can cite the fact next to the opinion.

Which platforms

The citation source list decides. As a rule of thumb:

  • Consumer brands: Trustpilot, Google Business Profile and the review sites of your own market and category.
  • B2B software: G2 and the platforms it now owns, plus Gartner Peer Insights. In G2's own data, review platforms account for 13.2% of AI citations at the evaluation stage, against 7.4% at discovery.
  • Local businesses: Google Business Profile first, then maps and directories fed from it.

Claim every profile, including the ones you never asked for. An unclaimed profile with a low score still gets cited.

Replies are content

Your reply to a review is free text that a model reads. A reply that names the problem, states what changed and gives a date is evidence. A copied apology is not. We write reply guidelines with your service team and review the first hundred replies together.

What we do not do

  • No fake, bought or incentivised reviews. Fake reviews are banned under EU consumer law.
  • No review gating. We do not ask only happy customers.
  • No AI-generated replies posted unread. A person reads and approves every reply.

How we measure it

Score, volume and recency per platform. Share of reviews with a reply. Number of pages with machine-readable reviews. In sampled answers: how often a review source is cited next to your brand, and the sentiment of the answer.

Which review platform matters most for AI answers?

The one the engines cite in your category and country. For consumer brands that is often Trustpilot or Google. For B2B software it is often G2. Your sampled answers show it within a week. Spread matters more than the single best platform.

Can we show review stars in Google for our own reviews?

Not on your organisation or local business pages. Google treats reviews a business controls about itself as self-serving and does not show stars for them. Product pages with genuine product reviews can qualify.

How do we deal with negative reviews in AI answers?

Fix the cause, reply in public with what changed, and collect new reviews from current customers. Models weigh recency. An old complaint with a dated solution and fifty newer reviews reads differently from an open one.

How many reviews do we need?

There is no threshold. A model looks at the score, the count, the dates and the text together. A steady flow of new reviews on two or three platforms beats a large old pile on one.

See which sources carry your category

The GEO baseline shows which pages the engines cite for your prompts, and where your brand is missing.