When someone asks ChatGPT to recommend a plumber, the AI is not sorting by 4.8 vs 4.6 stars. It's reading the words inside your reviews — the specific services mentioned, the problems described, the details customers volunteered. A business with 80 reviews and vivid, specific language will almost always outrank a competitor with 200 reviews that all say "great service, highly recommend."
The star rating is a weak signal to an AI. Review text is a rich signal. Here's what that means in practice.
Why AI Reads Reviews at All
AI assistants recommend businesses by matching your profile to the query. If someone asks "who does emergency furnace repair in Columbus," the AI is looking for evidence — across your Google Business Profile, your website, and your reviews — that you do emergency furnace repair in Columbus.
Reviews are one of the few places customers describe your business in their own words. Those words often match the exact phrasing future customers use when asking AI for help. A review that says "they came out at midnight when our heat stopped working" is more useful to an AI than your own homepage's "24/7 emergency HVAC services," because the review is a third-party, natural-language confirmation of what you do.
This is a big part of why AI gets your business info wrong — not enough third-party sources saying the same thing in plain, specific terms.
What Review Text Actually Contains (That Stars Don't)
Think about what a specific review tells an AI:
- Services performed: "She cut and highlighted my daughter's hair for prom."
- Problem solved: "Our drain was completely blocked — they cleared it in under an hour."
- Neighborhood or context: "Closest bakery to the Lincoln Park farmer's market."
- Who the customer was: "As a first-time homebuyer, I had no idea what to ask."
- Price range: "Cheaper than the downtown spots, similar quality."
- Timing: "Got us in same-day, no waiting weeks for an appointment."
Each of those details is a data point an AI can use to match your business against a query. A dental office with reviews mentioning "nervous patients," "kids," "no-wait appointments," and "accepts Medicaid" will get surfaced for a much wider range of queries than one whose reviews only say "friendly staff."
The Specific Words That Do the Work
AI recommendation systems pay attention to nouns and noun phrases — the specific things reviews name. Generic praise like "amazing," "wonderful," and "five stars" carries almost no information. It doesn't help the AI decide whether you're right for this particular question.
What helps:
- Service names: "root canal," "deck refinishing," "sourdough loaves," "prenatal massage"
- Conditions or situations: "leaking pipe," "first-time filer," "adopted rescue dog"
- Outcomes: "pain-free," "finished on time," "under budget," "still holding up two years later"
- Comparisons: "better than the chain salon I used to go to"
A bakery in Austin that has reviews mentioning "gluten-free birthday cakes," "custom wedding cakes," "nut-free," and "vegan options" is going to get recommended for far more specific queries than a competing bakery whose reviews just mention "delicious" and "fresh."
How to Influence This Without Buying Fake Reviews
You cannot write reviews for yourself, and you should not try. But you can influence what reviewers mention.
Ask specific questions when requesting a review. Instead of "please leave us a Google review," try: "Would you mind mentioning what you came in for and how it went?" That gentle prompt shifts reviews from generic to specific.
Respond to reviews and mirror the language. When you respond to a review that says "fixed our water heater on a Sunday," your response that echoes "glad we could handle your Sunday water heater repair" reinforces the same terms in a publicly crawlable block of text.
Flag gaps in your review coverage. If you offer eight services but reviews only ever mention two of them, AI will underweight the other six. Check which services have zero review mentions and target those when asking satisfied customers for feedback.
Do not ignore negative reviews. AI doesn't filter them out. A response to a negative review is still a text block associated with your business — keep the language professional and specific to the service involved.
Recency Matters, But Not How You Might Think
A common question is whether newer reviews outweigh old ones. For traditional search rankings, yes, recency is a factor. For AI, it's more nuanced. AI models are trained on data snapshots, not live feeds. A review from three years ago that vividly describes your water damage restoration service may still be in the training data that shapes how the AI describes you today.
What recency does help with is Perplexity and other AI tools that browse live search results. For those, a steady flow of recent reviews keeps your profile appearing fresh in the sources those tools pull from. Stopping review generation for a year is a real gap — not because of a recency algorithm, but because your profile looks inactive to crawlers.
See the full picture on how to get recommended by ChatGPT — reviews are one piece, but they work together with your schema markup, your website copy, and your business description.
What to Do This Week
Run a free AI visibility scan on your business, then open your most recent 20 reviews and read them for specificity. Count how many mention a real service by name versus how many just say "great experience." That ratio is a rough proxy for how much useful signal your review profile is sending to AI.
If the count of specific mentions is low, the fix isn't buying reviews — it's changing what you ask for when you invite happy customers to share their experience.