Why reviews weigh so much
Three effects stack up, and they do not replace one another.
Local ranking. Volume, freshness and average rating are among the signals that decide your place in the map results block. An equivalent competitor with forty recent reviews goes ahead of you and your six reviews from last year.
The click. Even ranked well, a listing without reviews loses to the one next to it. That is a human, immediate judgement that search optimisation does not undo.
Citation by AI. This is the new stake, and it is poorly understood. When someone asks an assistant to recommend a tradesperson, the model looks for elements that are verifiable and attributable. A structured rating is a fact it can reuse; a wall of free-text testimonials on your website is not.
The method: three decisions
When to ask. Response rates collapse over time. The useful window sits just after the moment the customer saw the result: the end of the job, the delivery, the first week of use. Not on invoice day, which ties the request to a payment.
How to word it. Three rules change everything:
- Ask a person, not a list. An individual, signed message that recalls what was done.
- Do not suggest the rating. « If you were happy, give us five stars » is steering, and steering is precisely what the regulation objects to.
- Ask an open question rather than requesting a review. « What helped you most? » produces useful text; « could you leave us a review? » produces « Great, thanks ».
Where to send. A direct link to the review form, not to your listing. Every extra click costs a share of the answers. Your business listing provides that short link, ready to paste into a text message or an email.
What the law says, and what it forbids you
Across the European Union, publishing or buying fake reviews is a misleading commercial practice. Since the directive modernising consumer law, the ban is explicit: it covers fake reviews, but also fake endorsements and presenting reviews as verified when nothing guarantees it.
Two practical consequences, often overlooked:
- Filtering out unhappy customers is risky. The trick of surveying the customer first, then sending the public review link only to the satisfied ones, manufactures an unrepresentative rating. That is exactly the behaviour the text targets.
- Paying for a review must be disclosed. Any consideration, even symbolic, must be flagged if the review is published.
This is not only a legal question. A perfect score across a hundred reviews inspires less trust than a 4.6 with a few reservations: buyers read the negative reviews first.
Replying matters as much as collecting
A public reply is not addressed to the person who wrote the review, it is addressed to the next readers. A negative review you answer calmly, naming what happened and what you did about it, converts better than a spotless profile.
Three principles are enough: reply to everyone, positive reviews included, in a few lines; never argue the facts in public when the customer is wrong, but offer to pick it up directly; and do not copy and paste the same formula, which shows immediately.
For the rest of the local mechanics, details, service area and markup, everything is in the local SEO guide.
How many reviews, and at what pace
The question always comes up, and the honest answer is that no published threshold exists. What is observed instead is that pace matters more than total.
Thirty reviews spread over three years read as continuous activity. Thirty reviews posted over two weeks, then nothing for a year, read as a campaign, and that is exactly the pattern platforms try to spot. A working business receives reviews all year, at an irregular but never zero pace.
A workable target for a small services business: ask systematically, at the end of every job, with no campaign. The real response rate often sits between one customer in five and one in three when the request is individual and well timed. A business doing ten jobs a month thus gets two to three reviews monthly, which is plenty.
Two useful side effects. First, reviews age: platforms and buyers give more weight to recent ones, so an old stock loses value without you doing anything. Second, the first negative review does more damage on a profile of six than on a profile of sixty: volume is also a shock absorber.
Finally, do not concentrate everything in one place. A solid profile on the main platform plus a few reviews on a trade directory beats a single collection point, if only because an account can be suspended.
What our audit measures
Two checks bear directly on reviews. In the AI optimisation category, the « reviews and ratings declared » check distinguishes three states: a rating declared as structured data, reviews visible but not declared, or no review detected. That declaration is what lets Google show stars and an AI cite your rating.
The business listing and local SEO category, holding 8 checks, looks at the remaining proximity signals. It only appears for websites presenting themselves as a local shop or service: a report without that category does not mean it was forgotten, it means it does not apply.
What the audit does not do: it does not read your business listing, does not count your reviews and does not judge their authenticity. It measures what your website declares. The volume and quality of reviews are checked directly on the platforms.