For twenty years the answer to "how do I rank" was links. Get mentioned, get linked, climb the list. An entire industry grew up around acquiring them, and a smaller, shadier industry grew up around selling them.
That game is not dead, but it has been demoted. Something else now decides whether a buyer ever hears your name.
When someone asks an assistant who they should hire, the answer is assembled from what other people have said about you. That means reviews, forum threads, roundups, comparison articles, directory listings, even a Reddit comment from 2024.
Your website is one voice in that assessment. Everyone else is the rest of the choir.
So reviews are picking up the job backlinks used to do. Google already says so for local search: "More reviews and positive ratings can help your business's local ranking." An assistant goes a step further and reads the words. Detailed reviews on several platforms are one of the strongest visibility levers a small business has.
It is how a business with a mediocre website and 200 detailed reviews can beat a business with a beautiful website and eleven.
Why third party sources carry so much weight
Put yourself in the model's position. You have two sources describing the same business. One is the business itself, saying it is the leading provider of quality solutions. The other is forty seven customers describing the jobs they had done and how each one turned out.
Only the second is any use for making a recommendation.
Self description is where marketing lives, so a careful reader discounts it, and a system built to give reliable answers has every reason to do the same. Independent corroboration is what moves a business from a name in the index to a name in the answer.
Links became valuable for the same reason. A link was a vote from someone else. A detailed review is a vote with reasons attached, and an assistant needs reasons to explain why it's recommending you. Ranking well on Google no longer guarantees you a place in the answer, a shift we unpack in why ranking number one no longer gets you into the answer.
On your website you control the message. In the AI answer, the internet's consensus about you is the message. The only question is whether you have shaped it.
The four sources that get read
Reviews with detail. The text matters more than the stars. "Five stars, great job" contributes almost nothing, because it contains no fact. Compare it with "They rewired a 1970s cottage in Umina in three days, worked around our tenants and came in eighty dollars under quote." That one names the service, the suburb, the timeframe, the constraint and the outcome.
Roundups and listicles. Third party "best of" articles suit assistants because they are already structured as a ranked recommendation. When Peec AI analysed nearly 200,000 AI responses, B2B software brands ranked first in a frequently cited listicle showed up in AI answers 16.5 percentage points more often, the biggest gap of the three industries it studied. That is a correlation in software and finance, not a promise for a plumber, but it tells you where these systems look.
Forums and communities. Reddit threads, trade forums and question and answer sites read as unfiltered peer opinion, which is the kind of text an assistant can quote.
Industry and supplier mentions. Being listed as an authorised installer or a member of your trade association. Dull, and exactly the kind of corroboration a machine can check.
Between them, these sources tell a model you're a real, established business worth naming. For the bigger picture, read how to make AI trust your brand.
How to engineer detailed reviews instead of star ratings
"Please leave us a review" gets you five stars and four words. Change the ask and you change what comes back.
- Ask a question. "If you have two minutes, what were you worried about before you booked us, and what happened?" A question like that invites three or four sentences with real facts in them.
- Ask within 48 hours of the result. Detail fades fast. A week later you get generic gratitude.
- Make the specifics easy. Remind them what you did. "Following the roof restoration in Terrigal last Thursday" hands them the details without you writing the review for them.
- Spread across platforms. Google, industry directories, Facebook, and any marketplace relevant to your category. Four sources that agree are harder to doubt than one profile.
- Reply to every one, with substance. Your replies are public too, and each one adds another paragraph of specific text about your business.
Build the review request into your process
Good intentions produce a burst of reviews and then nothing. A system produces them every week. Set one up like this:
- Pick the trigger moment. Choose the point where the result is fresh, such as the day the job is finished or the product has been delivered and used. Put the review request there for every customer.
- Write the message once. A short text or email using the question-based ask above, with the job details filled in and a direct link to the review page.
- Automate the send. Many CRMs and booking systems can send it when a job is marked complete.
- Rotate the platform. Send some customers to Google, some to an industry directory and some to Facebook, so your coverage grows beyond one profile.
- Follow up once. A single polite reminder a few days later. No more.
- Review the results monthly. Read the new reviews. Are they detailed? Do they mention services and locations? Adjust the prompt if they're thin.
One rule isn't negotiable. Never buy reviews or write them yourself. On Google you cannot reward them either: its policy bans incentives "in exchange for posting any review", and it also bans asking only your happy customers. In Australia the ACCC says fake reviews are against the law. A profile that looks manufactured also wrecks the trust you're trying to build.
Getting onto the lists
Most business owners feel awkward about this part, and they shouldn't. A "best of" list with four entries is a bad article, so the people who publish roundups need businesses to include.
Find the roundups that already rank for your category and city. Read what the included businesses have in common. Then email the publisher with what they need to add you: a clear description, real credentials, a photo, a specific differentiator and a customer they can contact.
Do that ten times and you may well land a few. A few third party listings can be a bigger visibility win than a year of blog posts on your own domain, because somebody else is vouching for you.
The consistency rule that ties it together
One last thing, and it undoes a lot of good work when ignored.
Every source describing you needs to agree on your business name, service area, core services and phone number. When a model finds three different service areas across four sources, it has no clean answer to give, and naming someone else becomes the safer option.
Spend an afternoon auditing every profile and directory listing you can find with your name on it. Fix the stale ones. Kill the duplicates. It is boring work, and we would rank it above most campaigns.
Handling the bad one
You will get a bad review. Everyone does, and the reply is public for as long as the review is.
Reply quickly and specifically, and leave the defensiveness out. Acknowledge what happened, say what you did about it, and offer to resolve it. Never argue the facts in public, even when you are right.
A thoughtful reply to a two star review is often more persuasive than the five star ones above it. It is the only evidence a prospect will ever get about how you behave when a job goes wrong.
A profile with a handful of imperfect reviews also reads as real, and a wall of flawless five stars makes people suspicious.
A worked example: the physio with the thin profile
Here's a hypothetical. Say you run a physiotherapy clinic in Newcastle with a strong website, a decent Google ranking, and thirty reviews that mostly say "great service, highly recommend." When you ask ChatGPT for a good physio in your suburb for running injuries, a competitor with a plainer website gets named. Their reviews talk about shin splints, marathon training plans and getting back to parkrun.
Over the next quarter, you change the ask. Every patient finishing a treatment plan gets a text asking what they came in with and what they can do now. Some go to Google and some to a health directory, and each review gets a reply with a line about the treatment approach. Two local running clubs publish recommended provider lists, so you email both. An old directory listing still shows your previous address, so you fix it.
Nothing on the website changed. But there is now a growing body of independent text connecting your name to running injuries in your suburb. You'd watch how specific the new reviews are, how many platforms carry them, and whether assistants start naming you when you test the same questions each month.
How to measure whether it's working
- Test the prompts monthly. Ask ChatGPT, Perplexity and Gemini the questions your customers ask. Record whether you appear and how you're described. Our guide to what AI is saying when customers ask about you covers how to do this properly.
- Track review volume and detail across platforms. Length and specificity tell you more than the star average.
- Count third party listings. Roundups, directories and partner pages that mention you.
- Watch AI referrals and branded search. More recommendations should show up as more people arriving from assistants or searching your name. We explain how to see that traffic in why traffic from AI converts better.
The reframe
Stop thinking of reviews as reputation management, a defensive chore handled when something goes wrong.
They are distribution now. The machine builds its sentence about you out of them, and that sentence decides whether a buyer ever types your name.
Forty detailed reviews is a marketing asset. Four hundred is a moat. Our digital marketing team builds the review systems, listing outreach and consistency audits that grow that asset month after month.
Frequently asked questions
Do reviews help you show up in ChatGPT and other AI answers?
They can. AI assistants build recommendations from what third parties say about a business, and reviews are one of the richest sources available. A detailed review that names the service, the suburb and the outcome gives the model a concrete reason to recommend you, where a star rating alone gives it very little. Reviews on several platforms also give it more to cross-check than a big number on one profile.
How do I get customers to leave more detailed reviews?
Ask a question. Something like "what were you worried about before you booked us, and what happened" prompts specific, useful answers. Send the request within 48 hours of the result, remind the customer what you did for them, include a direct link, and follow up once. Automating the request through your booking or CRM system keeps reviews arriving every week.
Which review platforms matter most for AI search?
Google is usually the most important, but spread matters. Industry directories, Facebook, relevant marketplaces and trade association listings all give AI systems more independent sources to cross-check. A business described the same way on four platforms is easier to trust than one with all its reviews in one place. Make sure your name, service area and contact details match everywhere.
Should I reply to negative reviews?
Always, and quickly. Acknowledge the issue, explain what you did about it and offer to resolve it privately, without arguing the facts in public. Prospects read replies. A calm, specific response to a critical review often builds more trust than the positive reviews around it, because it shows how you behave when a job goes wrong.