Most owners think reviews are about one number: the star rating. AI sees a richer picture. A perfect 5.0 from four reviews three years ago tells a very different story than a steady 4.7 from two hundred reviews, ten of them this month — and the AI reads that difference more carefully than any customer ever would.
To understand why reviews carry so much weight, go back to what the assistant is really doing. It is about to vouch for a business to a stranger, and it cannot personally verify that the business is any good. Reviews are the closest thing it has to asking the neighbors. They're independent, hard to fake at scale, and rich with the kind of detail the AI can't get from a website the business wrote about itself. That's why, when an assistant is deciding whether to put a business on a two-name shortlist, its review profile is one of the loudest trust signals it has — it's outside corroboration, and outside corroboration is exactly what a cautious recommender trusts most. Three dimensions matter, and the star rating is only one of them.
A handful of reviews, even glowing ones, doesn't give the AI much to stand on — it could be luck, or friends and family, and the assistant knows it. Volume turns an impression into a pattern. Once there are dozens of reviews saying similar things, the signal becomes statistically meaningful and the AI can lean on it without much risk of being embarrassed. A business doesn't need thousands. It needs clearly more than "a few," and ideally enough to look credible next to the competitors the AI is weighing it against — because this is always a comparison, never an absolute bar.
This is the dimension businesses neglect most, and the one the AI cares about more than people expect. A pile of reviews that all stopped two years ago suggests a business that has slowed down, changed hands, or coasted — and recommending a business that used to be good is just another way for the AI to be wrong today. Fresh reviews prove the business is active and still delivering. A consistent trickle — a few every month — is far more valuable than a one-time burst followed by silence, because it keeps the reputation alive in the present tense, which is exactly the tense the AI is making its decision in. A burst of reviews in a single week, followed by nothing, can even read as suspicious — the steady, ordinary cadence of a real, busy business is the more convincing signal precisely because it's harder to fake.
The AI doesn't just count stars; it reads. Reviews that mention specific services, specific staff, and specific outcomes help the assistant match a business to specific requests. A review that says "fixed our burst pipe at 11pm" makes a plumber a stronger candidate for "emergency plumber" than ten that just say "great service." A client can't write their own reviews, but they can shape them — by asking happy customers, at the right moment, to mention what was actually done for them. That's a coachable habit, and a legitimate one.
This is also where reviews quietly reinforce the rest of the picture. The assistant is constantly cross-checking what it reads, and a body of reviews that names the same services and the same location the business profile already claims is one more independent source agreeing with everything else. Star rating gets the business considered; the language inside the reviews is often what decides which of two close candidates actually gets named.
There are no shortcuts here that survive contact with reality — fake reviews get detected and do real, lasting damage to the exact trust you're trying to build. The durable approach is simple and a little boring: ask every satisfied customer, make it easy with a direct link, ask at the moment they're happiest, and respond to the reviews that come in. Responding is itself a recency and activity signal, another small "we're paying attention" the AI quietly notices.
Reputation is the factor where an agency most needs to manage expectations, because it is the slowest to move and the most visible when it does. You cannot manufacture two hundred recent reviews in a week, and you shouldn't pretend to — the work is steady, and the payoff arrives gradually. That creates the classic retention trap: real progress is happening, but a client glancing at a single star rating sees almost nothing change month to month.
This is why reputation progress only really shows over time, and why week-over-week tracking of review growth and recency is what turns "we're working on it" into something a client can actually see. A rating that ticks from 4.4 to 4.6 looks like noise in isolation; a chart of new reviews arriving every single week, with the freshness trend climbing, looks like momentum — and momentum is what keeps a client patient while the shortlist slowly rearranges in their favor. Sell the trend, not the snapshot, and reputation becomes the most durable story in the whole engagement.