How online reviews shape AI recommendations (and what to do about yours)

When someone asks ChatGPT or Perplexity for "the best physio in Newcastle" or "a reliable electrician near Geelong", the answer is not pulled from thin air. The model is weighing evidence from across the web, and your reviews are some of the strongest evidence it has.

Most owners still think of reviews as social proof for humans: stars that make a hesitant buyer feel safe. That job has not gone away: BrightLocal's long-running consumer survey finds 97% of consumers read reviews for local businesses. But reviews now do a second job, and it may matter more over time. They tell AI assistants whether your business is real, active, liked, and what it is actually known for.

What AI assistants take from your reviews

Generative engines do not just count stars. They read. A review profile gives a model several distinct signals:

• Legitimacy. A business with a steady history of reviews from real-sounding customers is far safer to recommend than one with three reviews from 2021. Models are tuned to avoid embarrassing recommendations, and a thin review profile reads as risk.
• Specifics to cite. When a dozen reviews mention "same-day hot water repairs", an assistant has a concrete, citable claim to repeat. Vague praise ("great service!") gives it nothing to work with.
• Recency. A profile where the most recent review is eighteen months old suggests a business that may no longer trade. Recency is one of the cheapest signals to maintain and one of the most damaging to neglect.
• Consistency with everything else. If reviews describe a Brisbane plumber but your website says you are a "national maintenance solutions provider", the mismatch weakens the whole entity. Reviews, Google Business Profile, and your service pages should all describe the same business.

Where reviews matter most

Google reviews carry the most weight for local intent, because they sit on the profile AI systems consult first. But spread matters too. A profile that is strong on Google and visible on one or two industry-relevant platforms (a trades directory, a health booking platform, a travel site, whatever fits your category) looks like a business with a real footprint rather than a single curated page.

This is the same principle behind the local SEO basics: independent sources agreeing about who you are and what you do.

How to build a review profile that earns citations

The good news is that the right approach is also the simple, honest one.

• Ask consistently, not in bursts. A steady trickle of reviews beats a one-off campaign that produces twenty reviews in a week and then silence. Bursts look paid for, to humans and to models.
• Ask at the right moment. The best time is just after the job is done well: the invoice is paid, the problem is fixed, the customer is happy. Make it a step in your process, not an afterthought.
• Make it effortless. Send the direct review link. Every extra step costs you a percentage of willing reviewers.
• Nudge towards specifics. You cannot script reviews, but you can ask "would you mind mentioning what we helped you with?". Specific reviews are the ones that get quoted.
• Reply to everything, especially the bad ones. A calm, factual reply to a negative review is read by every future customer and every model that processes the page. It shows an active, accountable business. Silence and arguments both read poorly.

What not to do

Fake reviews, bought reviews, and review gating (only inviting happy customers to a public platform while diverting unhappy ones) all violate Google's review policies, and the ACCC is explicit that creating fake or misleading reviews, or arranging for others to do so, is against the law. The penalties are real, but the quieter cost is the one that matters here: platforms and models are getting better at detecting manufactured patterns, and a profile flagged as suspicious is worse than a modest, genuine one.

If your review profile is thin, the fix is patience and process, not shortcuts.

Online reviews and AI FAQs

Do AI assistants actually read my reviews?

Systems that ground their answers in live web data, like Perplexity, Google AI Overviews, and ChatGPT when it searches, retrieve and process review content alongside your website and business listings. Review themes regularly surface in how these tools describe a business. Even where a model is not quoting a specific review, the aggregate signals (volume, rating, recency) shape whether you appear at all.

How many reviews do I need?

There is no magic number, and it varies wildly by category and area. The more useful question is how you compare to the competitors an AI assistant would recommend instead of you. If the three other firms in your area have hundreds of recent reviews and you have nine, that gap is part of your visibility problem. A competitor scan makes the gap concrete.

Are Google reviews more important than other platforms?

For local commercial intent, yes. Your Google Business Profile is the single most consulted source for "near me" style questions. But industry-specific platforms matter in categories where buyers research there first, and a presence on more than one platform strengthens the legitimacy signal. Start with Google, then add the one or two platforms your customers actually use.

Should I respond to negative reviews?

Yes, and promptly. A good response acknowledges the issue, states the facts calmly, and offers to resolve it offline. What you are really doing is writing for the hundreds of future readers, human and machine, who will judge the exchange. Never argue, never reveal customer details, and never write while angry.

Is it legal to ask customers for reviews in Australia?

Asking for honest reviews is fine. What crosses the line is paying for reviews, offering incentives contingent on a positive review, writing reviews for your own business, or selectively gating reviews. The ACCC has published guidance on online reviews, and the safe rule is simple: every review should be a real customer's honest opinion, freely given.

My competitor has fake reviews. What can I do?

Report them through the platform's reporting tools and move on. Do not respond in kind. Manufactured profiles tend to decay (platforms purge them, patterns get flagged), while a genuine profile compounds. Your energy is better spent making your real customers' experience worth writing about.