AI search · Evidence guide

Do Google reviews affect AI recommendations?

Probably related, but unproven. One large study found that businesses AI tools mention have far more Google reviews. But that is correlation, not proof that reviews cause recommendations. No AI platform publishes a rule.

Updated October 2026

The short answer in more detail

There is evidence that Google reviews are relevant to how some AI search systems describe or recommend local businesses. The strongest published evidence here is observational: businesses with more reviews were often more visible in a study, and review information can appear in AI search contexts. That is different from proving that adding ten reviews causes a recommendation. A well-known business may have more reviews, a better website, wider directory coverage, and more local search visibility at the same time.

ChatGPT, Perplexity and Google AI Overviews are different products with different sources, and none has published a rule like 'reach 100 Google reviews and you rank'. The same tool can answer differently for 'the closest emergency plumber' and 'a highly rated dentist'.

The practical goal is accurate business information and genuine customer feedback where customers already look. Do not buy reviews or script what customers say. Google bans fake and incentivized reviews. A dishonest campaign also spoils the evidence you want AI tools to find.

Sources: Insites: AI visibility benchmark across 10,000 local businesses, BrightLocal: Uncovering ChatGPT Search Sources, Google Maps: prohibited and restricted content

What BrightLocal measured in ChatGPT Search

BrightLocal's 2024 research examined 800 manual local-business web searches in ChatGPT Search and recorded displayed sources. In that sample, business websites accounted for 58% of listed local search sources, business mentions 27%, and directories 15%. The study says ChatGPT can use review information from Google and Google Maps. But Google Maps did not appear as a directory source among the first displayed sources in its sample. Those statements can both be true: using a piece of information and listing a domain as a cited directory are different observations.

The study covered 20 US cities and 20 business categories, with two kinds of query, in November 2024. It did not test what happens when a single business gains reviews, nor did it establish a formula for how ChatGPT chooses a recommendation in 2026. Its source categories also mix websites that directly belong to businesses with other pages that do not fit the directory or mention definitions. Read the methods before turning the percentages into a plan.

The practical clue: your own site can be a source of business information. Keep service pages, contact details and location accurate, and make reviews inspectable at their original source. The citations in one ChatGPT answer are worth inspecting, but they are not a log of everything the model considered.

Sources: BrightLocal: Uncovering ChatGPT Search Sources, OpenAI: searching the web with ChatGPT

What Insites found across 10,000 businesses

Insites ran 10,000 US local businesses through a standard audit, then tested each one in ChatGPT and Perplexity. It separated awareness of a business from favorable descriptions and from actual recommendations for relevant local searches. In its analysis, businesses found by both systems averaged 133.4 Google reviews, while businesses found by neither averaged 10.7. It also reported that its most-recommended group averaged 304.2 Google reviews, against 117.5 for its least-recommended group.

Those are study-group averages, not a target or a threshold. The observed relationship may partly reflect business size, age, location, marketing activity, or reputation. The study also looked at Business Profile completion, local search presence, directory consistency, and website content. It cannot tell an individual shop how many new reviews will produce a particular AI answer. Insites sells AI visibility products, so its published method and findings are useful evidence while still deserving independent scrutiny.

Review volume appears alongside other signs of an established business. A current profile, accurate details and a useful site are within your control; an assistant's recommendation is not. If you already have many reviews but are described wrongly, fix citations and business details before chasing more reviews.

Sources: Insites: AI visibility benchmark across 10,000 local businesses, Insites: correlation study of reviews and AI visibility

ChatGPT: citations are clues, not a ranking report

ChatGPT search can list the sources it used; BrightLocal's study recorded those links for every search. Like any AI answer, the result can be wrong or out of date. OpenAI's help says ChatGPT may use your approximate location for questions about nearby businesses. A business owner can ask several realistic customer questions, inspect the source links, and note factual errors. A single answer is a snapshot from one wording and one context, not a stable ranking position.

Try prompts customers would use, such as 'Which local electrician handles urgent calls in this town?'. Check whether the assistant names your business, cites the right site, and gets your service area and hours right. Repeat with another phrasing and neighborhood. A question that includes your own name is not an independent recommendation.

If a cited page is stale, correct that source. Also check that your site does not block OAI-SearchBot (OpenAI's search crawler, the bot that reads web pages). OpenAI's crawler guide says sites that block it will not be shown in ChatGPT search answers. Allowing it does not guarantee you will be shown or recommended. A review page that clearly identifies your business and links to the original Google profile gives readers a way to verify the content. That is true whether or not an assistant cites it.

Sources: OpenAI: searching the web with ChatGPT, OpenAI: overview of OpenAI crawlers, BrightLocal: Uncovering ChatGPT Search Sources

Perplexity: inspect the actual source links

Perplexity's help page on Pro Search says every answer includes direct links to the original sources. It searches articles, forums, videos and other pages, depending on the query. That makes a source inspection useful. If an answer recommends a rival, look at what it cited. Then check whether your business has a comparable public page with the relevant information. A missing citation is a clue to investigate, not proof that a crawler has rejected your site.

The Insites study included Perplexity alongside ChatGPT, but it did not publish a universal Perplexity review-weight rule. Avoid transplanting a finding from a combined study into a claim about Perplexity's private scoring. A query about a product type may draw different pages from a query about local availability. Customer reviews can supply evidence of experience, while your own site should provide concrete services, location, prices where appropriate, and contact details.

If an answer cites a directory or old article with a wrong phone number, correct the source you can influence. Also make your own contact information easy to find. Do not manufacture reviews to alter the answer. Real customers' feedback is valuable because it is independent of your own sales copy, and platforms have policies against manipulation.

Sources: Perplexity: how Pro Search uses sources, Insites: AI visibility benchmark across 10,000 local businesses, Google Maps: prohibited and restricted content

Google AI Overviews: different rules, less visibility into weighting

Google's Search Central documentation says a page needs no extra technical requirements to be a supporting link in AI Overviews or AI Mode. It only needs normal Search eligibility (indexed and able to show a snippet, a short text preview). Inclusion is never guaranteed. Google advises accurate Business Profile information and important content as text.

That guidance does not say embedded Google reviews cause an AI Overview recommendation, or give a review-weight formula. A Business Profile rating in a local result is a different thing from your website being cited in an AI Overview.

Google explicitly says you do not need a new AI text file, special schema (labels in your page code), or other AI-specific markup. You can appear in its AI features without them. This matters when evaluating llms.txt or review schema sales pitches. Good site basics and accurate visible information remain useful, but they are not an AI Overview switch.

Sources: Google Search Central: AI features and your website, Google Search Central: self-serving review snippets

What a hosted reviews page and llms.txt can actually do

VerifiedWidget can publish a hosted public reviews page for a widget. It also publishes an llms.txt file (a plain text file written for AI tools) for that page. The practical purpose is readability: a text-based page gives tools and people a place to inspect review content and source context. It is hosted on VerifiedWidget rather than your own domain. If you limit a widget to your own domains, the hosted page is hidden from search engines and its llms.txt file is switched off. You should link to it where useful and keep the underlying reviews accurate. But it does not transfer Google's rating to your site. It also cannot force any assistant to cite it.

Insites examined technical AI-readiness signals and found only a weak association for llms.txt in its 10,000-business study. That is not a randomized test of every possible implementation, but it is a reason to avoid selling llms.txt as an AI ranking lever. Google Search Central says no AI text file is required for AI Overviews. A readable page can still be useful because it exposes content in a direct format; the outcome should be described in that modest way.

If you are considering any review page, inspect the actual published text. Can a visitor identify the business and location? Are review authors, dates, and source links clear? Is the page current? A file alone cannot make an inaccurate or thin page trustworthy. Your main website should still explain your real services in your own words.

Sources: Insites: correlation study of reviews and AI visibility, Google Search Central: AI features and your website

A measured plan for a local business

Confirm your business name, address, phone, hours, categories and services on your Business Profile and website. Ask real customers for honest reviews with Google's link or QR code, with no rewards or preferred wording. Build pages that answer questions people ask before buying, and keep directory listings you control consistent.

Then test a few real customer prompts in ChatGPT, Perplexity and Google Search. Record the date, exact query, businesses named, errors and links shown, and repeat after you fix something. This is a practical audit, not an experiment.

Track what you can observe: correct mentions, referrals and contact requests. Do not report a fixed 'AI rank' from one answer. No study reviewed here proves a guaranteed recommendation in any AI product.

Sources: Google Business Profile: request reviews with a link, Google Maps: prohibited and restricted content, OpenAI: searching the web with ChatGPT, Google Search Central: AI features and your website

Frequently asked questions

Does ChatGPT use Google reviews?

BrightLocal's study says ChatGPT can use review information from Google and Google Maps. OpenAI has no public rule that ties review count to recommendations.

What did Insites find?

In its 10,000-business sample, businesses found by both ChatGPT and Perplexity had far more Google reviews on average than those found by neither. That is correlation, not proof of cause.

Will llms.txt make an AI assistant recommend me?

No. It can make information easier to read, but no recommendation is guaranteed. Google says AI Overviews do not require an AI text file.

Do Google AI Overviews use the same signals as ChatGPT?

No published evidence establishes the same weighting. Google describes separate AI Search features and does not publish a review-weight formula.

What should I do first?

Keep your Business Profile and site accurate, ask real customers for honest reviews, and inspect citations in sample answers.

Make your real reviews easier to read

Publish a clear review display and hosted page, while keeping expectations about AI recommendations honest.

Start free Compare plans

Sources and further reading

Want to continue? Browse installation guides.