Quick answer: Yes — as a qualifying filter. In SOCi's 2026 Local Visibility Index (~350,000 chain-brand locations, January 2026), ChatGPT recommended just 1.2% of them, and the ones it named averaged 4.3 stars. Reviews look like a bar you clear to be considered. That sample is multi-location brands, so a solo agent should read it directionally.
An agent I work with in Sonoma County has 28 Google reviews, every one of them five stars. When we ran the discovery queries in June, she was absent from every "best agent in Santa Rosa" answer on both ChatGPT and Perplexity. The single query where she did surface, ChatGPT credited a stranger's comment on Reddit.
Do AI assistants actually read reviews?
Yes, and some of them lean on reviews for nearly every recommendation. In an August 2025 study by Feefo, a review-software vendor, running 250 identical travel queries across ChatGPT, Gemini and Perplexity, Perplexity referenced reviews in 100% of its answers and ChatGPT in 58% — and purchase-intent queries pulled reviews 67% of the time versus 33% for informational ones. That's travel, and Feefo sells review software. Take the mechanism; the numbers are theirs.
The mechanism holds up against the local data. When a model has to put a person's name in an answer, it reaches for corroboration someone else wrote. SOCi's 2026 index found the sources large language models cite most on local queries are Google Maps at 32.5% and niche directory sites at 26.3%, with brand websites third at 23.1%.
How many reviews and what star rating do you need?
The working bar in 2026 is roughly 20 or more reviews at 4.0 stars and up — and the locations AI actually named averaged 4.3. BrightLocal's Local Consumer Review Survey 2026 (1,002 US adults, February 2026) found 47% of consumers won't use a business with fewer than 20 reviews, and 68% will only use one rated four stars or higher, up from 55% a year earlier. On the machine side, SOCi found ChatGPT's recommended locations averaged 4.3 stars, Perplexity's 4.1 and Gemini's 3.9.
| The bar | Human buyers (BrightLocal, Feb 2026) | AI recommendations (SOCi, Jan 2026) |
|---|---|---|
| Star rating | 68% require 4.0★ or higher; 31% require 4.5★+ | ChatGPT's picks averaged 4.3★ |
| Review count | 47% skip a business under 20 reviews | Not disclosed |
| Recency | 74% want reviews from the last 90 days | Not disclosed |
| Owner replies | 80% prefer a business that replies to all reviews | Not measured as a standalone AI signal |
One caveat on the SOCi column: that study measured multi-location brands with 50 or more locations. Read the bar directionally — the decimals are for chains.
Why do agents with great reviews still get left out?
Because the reviews live somewhere the machine can't connect to their name. In our 2026 pilot, only 2 of 18 fully-scored agents — 11% — exposed their reviews in a form machines can read and attribute to them. Across 12 California metros, none of the 55 agents we tracked appeared when we asked for the best agent in their own market, and every one of them sells $31–40 million a year.
Production and AI visibility turn out to have nothing to do with each other. You can be the busiest agent in your zip code, hold a wall of five-star reviews, and still be missing from the answer a buyer now trusts, because nothing in the machine's reach ties that 5.0 to you.
Do reviews on your own website count?
They give AI something readable, and they won't put star ratings on your own pages in Google's results. Since Google's September 2019 review-snippet update, reviews about your own business placed on your own site under LocalBusiness or Organization markup are ineligible for the star rich result — no manual action, Google just doesn't display them. That rule is still in Google's 2026 documentation, and plenty of agents get sold "review schema" without hearing it.
Mirroring your reviews onto a site you own still does real work, as long as they're visible text on the page: the reviewer's name and the date. Testing by the SEO firm searchVIU in October 2025 found AI engines pull from raw HTML and largely ignore JSON-LD and JavaScript-injected content, so a review that lives only in schema, or only appears after a script runs, may never be read at all. Keep the schema for Google's other features and put the words on the page for the models. Your Google Business Profile stays the source of record — it's what Google's AI surfaces and ChatGPT's licensed data partners draw from, which I walked through in how agents get recommended by ChatGPT.
How recent do your reviews have to be?
Recent enough that your newest one landed inside the last 90 days. BrightLocal's 2026 survey found 74% of consumers look for reviews written in the last three months, and 32% want them from the last two weeks. The engines don't publish how they weight review recency, so I won't pretend to know. What I can tell you is the human bar moved: a review profile that went quiet two years ago reads like a business that went quiet two years ago.
Volume follows from asking. On the case study, Andrew went from 27 five-star reviews at launch to 50 by late August by asking his own past and closing clients directly.
What should you do about your reviews this week?
Two things, and neither takes an afternoon. Open your Google Business Profile and reply to every review you haven't answered — 80% of consumers in BrightLocal's February 2026 survey say they're more likely to use a business that replies to all of them, and templated replies put off 50%, so write four real sentences. Then text the last three clients you closed with and ask, specifically, for a Google review that names the town and the kind of transaction.
After that, put the reviews you already have on a page you own, as visible text with names and dates, and the machine has something to read the next time someone asks it who to hire.
Frequently asked questions
How many Google reviews does a real estate agent need?
Twenty is the practical floor: 47% of consumers in BrightLocal's February 2026 survey won't use a business with fewer than 20 reviews, and 68% require a 4.0-star average or better. Recent and specific beats a big old pile.
Do Zillow or Realtor.com reviews help with AI search?
They help when an engine can find them and connect them to you — niche directory sites were the second-most-cited source on local queries in SOCi's 2026 index, at 26.3%. In our 2026 pilot, most agents' reputations sat on third-party profiles the models didn't reliably credit back to them.
Should you respond to every Google review?
Yes. BrightLocal's 2026 survey found 80% of consumers are more likely to use a business that replies to all its reviews, while templated replies put off 50% of them. Replies also put fresh, specific text on the profile.
Can more reviews get ChatGPT to recommend me?
More reviews can make you eligible for the answer; they don't buy you a place in it. SOCi's 2026 index found reviews and ratings "function less as a ranking factor and more as a filter for inclusion" — clearing the bar is what gets you considered, and the rest of your visibility, like whether your site is readable and your identity consistent, is what the engines are working from. Anyone promising you an AI recommendation for a review count is selling something.
Run the query yourself this week: ask an AI assistant for the best agent in your town and see whether you're in the answer. Get your free AI Visibility Score if you'd rather have it scored properly, including whether your reviews are in a form the machine can read. The mechanics live in the AI visibility FAQ.
Sources: SOCi 2026 Local Visibility Index (~350,000 multi-location brand locations, January 2026). BrightLocal Local Consumer Review Survey 2026 (1,002 US adults, February 2026). Feefo, August 2025, 250 travel queries across ChatGPT, Gemini and Perplexity — Feefo sells review software. Google Search Central review-snippet guidelines, September 2019 update, confirmed live in the 2026 documentation. searchVIU raw-HTML testing, October 2025. Own-data figures from the LeadJens 2026 AI Visibility Index pilot.