online reviews

Reviews in AI Search: What Trustpilot's Data Means for You

Trustpilot's study found review citations in AI answers rise from 2% to 24% near purchase. See what it measured and a compliant plan for your reviews.

A customer who is ready to hire someone may no longer start with a list of blue links. They may type a question into ChatGPT or Google's AI and ask which business to choose, and the answer can name a short list on the spot. On September 30, 2026, the retail press reported that reviews count for more in those answers as a buyer gets closer to a decision. The data behind that report comes from Trustpilot, and it is worth reading closely before you change anything. This post covers what the study found, which review sites showed up, and what you can do with your own reviews this month. You can run the first check in about ten minutes, and the review plan below stays inside the rules that apply to every business. Nothing here promises that a review count will get your business named, because no study supports that promise.

Key Takeaways

Reviews get cited more near purchase

In Trustpilot's study, review and trust sites made up 2% of cited sources for awareness prompts, 16% for evaluation prompts and 24% for purchase-intent prompts.

Check your own answers first

Ask ChatGPT and Google AI Mode the question a buyer would ask, note who is named and which review sites are cited, and count your reviews from the last 90 days.

Ask everyone, the same way, and never buy or filter

The FTC's 2024 rule, Google and Trustpilot all prohibit buying reviews and steering who gets asked.

Review citations rose from 2% to 24% as buyers got closer to choosing

The numbers come from a 2026 white paper called "What AI says about you," which Trustpilot published. Trustpilot says it commissioned SEER Interactive to analyze 804,491 AI responses about 1,926 US brands across eight industries. The full report says the testing happened in March 2026 across ChatGPT, Google Gemini, Perplexity and Google AI Overviews. Because Trustpilot is a review platform, this is research from a company with a stake in how reviews are seen, and it is best read as one dataset rather than a rule.

Across all the prompts, review and trust sites made up 14% of the sources the assistants cited. That share moved with the question being asked. Trustpilot's September 24, 2026 product release repeats the same finding: product reviews are cited more when customers are close to buying.

Review citations by buyer stage in Trustpilot's study
Buyer stageShare of cited sources that were review or trust sites
Awareness (learning what exists)2%
Evaluation (comparing options)16%
Purchase intent (ready to choose)24%
Buyer stageAwareness (learning what exists)
Share of cited sources that were review or trust sites2%
Buyer stageEvaluation (comparing options)
Share of cited sources that were review or trust sites16%
Buyer stagePurchase intent (ready to choose)
Share of cited sources that were review or trust sites24%

That is a count of citations, not a count of customers won. A business can be named in an answer and never get a call, and cited without getting a lead. Google describes AI Overviews as generated snapshots with links for going deeper, and OpenAI says ChatGPT Search answers can include citations and links but may be incomplete, outdated or incorrect. Your goal is to be described accurately when a buyer asks.

A closed laptop beside a stack of blank index cards and a pencil on a worn wooden desk.
The same question asked at three stages of a purchase can pull three different sets of sources.

Which review sites appeared depends on the assistant

Trustpilot's dataset ranked Trustpilot as the most-cited review platform among the platforms tested. The shares below are study-specific citation shares for each assistant, not permanent platform rules, and they will differ with the business category and the question.

Review platform citation shares in Trustpilot's March 2026 data
AssistantTrustpilotBBBG2CapterraYelp
ChatGPT58%15%7%3%0%
Google AI Mode21%4%5%3%0%
Gemini49%14%15%8%17%
Perplexity52%10%15%4%9%
AssistantChatGPT
Trustpilot58%
BBB15%
G27%
Capterra3%
Yelp0%
AssistantGoogle AI Mode
Trustpilot21%
BBB4%
G25%
Capterra3%
Yelp0%
AssistantGemini
Trustpilot49%
BBB14%
G215%
Capterra8%
Yelp17%
AssistantPerplexity
Trustpilot52%
BBB10%
G215%
Capterra4%
Yelp9%

Other vendors' studies report different source mixes, with heavier use of Reddit, YouTube or editorial sites, as in Conductor's comparison of how AI citations differ. So, a single site is not the answer for every business. Start with the platform where your customers already leave reviews, and keep your Google Business Profile accurate, since Google says Business Profiles can display trusted third-party reviews from travel or local review sites.

The study also looked at what the cited review text was about. The themes were customer service, shipping and delivery, product or service quality, whether the business is legitimate, and refunds or returns. Those are questions about how you treat customers, which a star rating alone does not answer.

You can run the check on your own business in ten minutes

Use one service, one town and one realistic buying question. Ask ChatGPT with web search turned on: "Which [service providers] in [town] should I consider for [specific need]? Compare recent customer feedback, experience, location, and any clearly stated pricing or availability." Then ask the same question in Google AI Mode.

  1. Record the answer: Write down the date, the businesses named, the URLs cited and whether yours appears.
  2. Open the cited review pages: See which review sites the assistants pointed to and what the recent reviews say.
  3. Count your recent reviews: Count the reviews dated within the last 90 days on Google and on any platform that appeared.
  4. Compare your details: Check that your business name, services, location, phone number, website and hours match across profiles.
  5. Repeat monthly: Use the same wording each time so you are comparing like with like.

Treat the result as one observation from one day. An answer can change with location, wording and the assistant, so a single check is not a stable ranking position.

A wooden desk with a blank notepad, a sharpened pencil and a small kitchen timer set beside them.
Keep the wording identical each month, and the results become comparable.

A compliant review plan asks every customer the same way

The FTC's Consumer Reviews and Testimonials rule took effect October 21, 2024. It prohibits buying or selling fake reviews, reviews from people without real experience, certain undisclosed insider reviews, review suppression, and incentives tied to a positive or negative sentiment. Platform rules add more detail. Google's contributor policy says reviews must reflect a genuine experience. It prohibits paid reviews, incentives, selective requests for positive reviews and requests for revision or removal in exchange for benefits.

Trustpilot's June 2026 business guidelines require invitations to be fair, neutral and unbiased, sent to everyone at the same point after the purchase, with no incentives. That includes customers who may have had a bad experience. Yelp goes further: its support page says businesses should not ask customers for Yelp reviews.

A monthly review routine

  1. 1

    Week 1

    Run the ten-minute check and record the result.

  2. 2

    Week 2

    Ask every customer who finished a job that month, using one neutral message.

  3. 3

    Week 3

    Reply to every new and unanswered review, without private customer details.

  4. 4

    Week 4

    Fix any recurring complaint in the reviews and confirm your profile details match.

In practice, each piece looks like this:

  • Ask everyone: Send the same request to every genuine customer after they have used the service, whether or not you expect praise.
  • Skip the wording: Let customers describe their experience in their own words. Do not script phrases or ask for keywords.
  • Offer nothing in return: No discounts, gifts or entries in a drawing for a review.
  • Reply to all of them: Answer positive and negative reviews without revealing private customer information.
  • Keep details consistent: Use the same name, address, phone number, categories and services on every profile.

For the full method of asking, see how to get more reviews, and for how steadiness and recent activity fit in, see our post on review recency and consistency.

A row of blank paper envelopes beside a rubber stamp and an ink pad on a clean counter.
One request, sent to every customer, is the routine that holds up under the rules.

No review count is a target, and no site guarantees a citation

Trustpilot reported median review counts of 13 and 81 for groups in its sample. Those figures describe its sample, not a threshold you need to hit. The study compared groups of brands and did not run an experiment, so it does not show that adding reviews causes an assistant to name your business. Review quality, relevance, location, business category, freshness and the assistant's own retrieval can all change an answer.

So, skip the magic number, skip the hunt to get every review onto every site, and skip asking customers to mention keywords. A real negative review with a professional reply tells a buyer more than a vague five-star rating. Spend your effort on accurate profiles, steady genuine feedback, replies, and fixing whatever customers keep complaining about.

A clean workbench with a tape measure, a level and a folded work apron on a wooden shelf.
Recurring complaints in reviews are usually a service problem first and a reputation problem second.

You can do this alone, and help makes sense when the check shows a wider gap

If your profiles are accurate, you have a working request routine, your recent reviews are genuine and you answer them, keep doing it yourself. Nobody can promise inclusion in ChatGPT, Google AI Overviews, Gemini or Perplexity for any fee, and an agency cannot control each assistant's model, retrieval or location settings.

Outside help earns its cost when the check points to a wider problem: inconsistent or duplicate profiles, unanswered reviews, recurring complaints, missing service or location detail, or no reliable way to compare AI answers with real leads and sales. In that case, the work is an audit, a compliant request process and monitoring, judged by inquiries and not by citation counts.

Check what you know about reviews in AI search

Pick an answer to begin.

1. In Trustpilot's study, what happened to review-site citations as prompts moved from awareness to purchase intent?

2. Which review request is compliant?

3. What does the study show about adding more reviews?

Do reviews matter in AI search?

They can matter more when a buyer is close to choosing. In Trustpilot's study, review and trust sites were 24% of cited sources at purchase intent, compared with 2% at awareness. No review count guarantees a recommendation.

Which review sites do AI assistants cite?

In Trustpilot's March 2026 sample, Trustpilot, BBB, G2, Capterra and Yelp all appeared, with shares that varied by assistant. Those shares are from one study and can change.

Should I buy reviews to show up in AI answers?

No. The FTC's rule prohibits buying or selling fake reviews, and Google and Trustpilot prohibit paid, biased or incentivized reviews.

Can I ask my customers for reviews?

Yes on Google and Trustpilot, if you send a neutral request to every genuine customer. Yelp says businesses should not ask customers for Yelp reviews.

What should I do about my reviews this month?

Run the same buying question in ChatGPT and Google AI Mode, count your reviews from the last 90 days, check that your profile details match, ask every eligible customer the same way, and reply to reviews already posted.

Will more reviews get my business cited?

No promise is justified. Trustpilot reported an association in its own dataset, not a rule that holds for every business or assistant.

What This Means for You

Reviews in AI search get cited more as buyers get closer to choosing, according to research Trustpilot commissioned from SEER Interactive. That raises the value of having recent, genuine reviews on the sites your customers use, and of replying to them. It does not turn a review count into a guarantee.

The practical move is small. Run the ten-minute check, compare your profile details, and set up one neutral request that goes to every customer. Reply to what is already there, and fix what customers keep mentioning. Repeat the check each month with the same wording, and compare the answers with the inquiries you actually receive.

If you want help with any of it, Web Leveling can audit your profiles and review routine through our online reputation service, and our generative engine optimization work covers how your business shows up in AI answers. If the check finds no real gap, we will tell you to keep it in-house. To talk it through, contact us about your reviews. We work with small and medium businesses across the country and overseas.

Terms

Review and AI search words in this post

Tap a term to see what it means.

AI Overview. A generated summary Google shows on some searches, with links to sources.

Citation. A source an AI answer points to as the place its information came from.

Purchase intent. A stage where a buyer is ready to choose a business or product.

Review gating. Steering who is asked for a review, such as only asking happy customers.

Review suppression. Blocking or removing reviews to hide what customers said.

Business Profile. A business's free listing on Google Search and Maps.