A referred buyer does not take the referral on trust. They check: a search of your name, your website, your LinkedIn, review sites, and sometimes an AI assistant. Make sure each one states the same facts, shows recent proof, and answers what you do and for whom in one read. Then check it monthly.
Referral-led businesses tend to underinvest here, for a sensible reason. The work has always come through people, so the website and the profiles feel like a formality. But a referral now arrives with a step in between. The person who was told your name goes and looks you up before they reply, and what they find either confirms the recommendation or quietly undoes it.
This note walks through the check a referred buyer runs, why consistent facts matter more than clever copy, how to see what AI assistants say about you, how to collect reviews without breaking the rules, and a 30-day checklist to fix what you find.
What does a referred buyer check before they reply?
The referral got you a look. The look decides whether you get a reply.
Picture the buyer. A peer said “talk to these people, they were good.” The buyer has your name and perhaps a first name. Before they write, they spend a few minutes confirming you are real, current and right for them.
In most cases the check covers some or all of these stops.
- A search of your company name, sometimes with the word “reviews”. They see the top results, any business profile, and whatever else ranks for your name.
- Your website. Mostly the homepage, the page for the service they were referred for, and the about or team page.
- LinkedIn. Your company page, and the profile of the person they were told to contact.
- Review and directory sites relevant to your field, if you appear on any.
- An AI assistant. A question such as “what does this company do, and what should I know before hiring them?”
What the buyer wants from all of it is simple. Is this company real and active? Does it do the thing I need? Has it done it for people like me? Is there anything worrying? None of that requires brilliant copy. It requires the answers to be easy to find and to agree with each other.
The note Your website went quiet: what buyers check before they reply looks at the website stop in detail. This note covers the rest of the route.
The goal is verification, not persuasion.
Why do consistent facts matter more than clever copy?
A buyer who finds three different descriptions of you starts wondering which one is true.
Referral-led firms accumulate drift. The website describes the business as it was three years ago. The LinkedIn page has an older tagline. A directory lists a former address. A founder’s profile still says a previous job title. Each is small. Together they make a careful buyer hesitate, and a careful buyer is exactly who a referral brings.
Drift also affects AI answers. An assistant answering from memory repeats what it learned during training, which may be old. An assistant searching the web reads whatever pages it finds, and if they disagree, the answer may pick either version or blend them. The note Remembered or retrieved: how AI assistants name companies explains the difference and why it changes how quickly a fix shows up.
The fix is a one-page fact sheet, written once and copied everywhere. It should hold these facts.
- Your name exactly as it should appear, and any former or trading names.
- One sentence on what you do, and for whom.
- The services you offer, named the way buyers name them.
- Where you work, and whether you serve clients remotely.
- When you started, and who leads the company, with their current titles.
- Three pieces of proof you can stand behind: named clients with permission, case studies, credentials, awards.
- How to get in touch, and what happens after someone does.
Then make every source match the sheet, word for word where you can. Consistency is what lets a buyer, a search engine and an AI assistant all reach the same conclusion about who you are.
What does an AI assistant say about you, and how do you check?
You cannot see the buyer’s conversation. You can have the same one yourself.
Open two or three assistants and ask what a referred buyer would ask. Ask each one twice: once with web search off, if the assistant allows it, and once with it on. Write down what comes back.
Five prompts cover most of what a referred buyer asks.
- What does [your company] do, and who is it for?
- What should I know before hiring [your company]?
- Who runs [your company]?
- What do clients say about [your company]?
- Who are the alternatives to [your company] for [your service]?
Read the answers for three things: facts that are wrong, proof that is missing, and the sources cited when search is on. A wrong fact usually traces back to a page somewhere, often an old directory listing or an out-of-date profile. Fix it at the source. Arguing with the assistant changes nothing for the next person who asks.
Two practical points from the platforms themselves. Google’s Search Central guidance on AI features, last updated in December 2025, says there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode beyond normal search good practice. And OpenAI’s crawler documentation says sites that block its OAI-SearchBot crawler will not be shown in ChatGPT search answers, which is separate from GPTBot, the crawler used for model training. Check your robots.txt file allows the search crawlers you want.
Treat each check as a spot check, not a ranking. Answers vary from run to run and by how a question is worded. The value is in noticing a wrong fact or a missing piece of proof, then checking again next month to see whether it changed.
How do you collect reviews without breaking the rules?
Ask everyone, the same way, and never pay for a particular opinion.
Referral-led firms often have many happy clients and few public reviews, because nobody asked. Asking is fine. How you ask is where the rules come in.
In the United States, the Federal Trade Commission’s Trade Regulation Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024. According to the FTC’s own questions-and-answers guidance, it bans fake reviews, including reviews invented with AI, and bans offering incentives in exchange for reviews that express a particular sentiment, positive or negative. It also bars suppressing negative reviews through threats, intimidation or false accusations, and expects company insiders who post reviews to disclose the connection clearly.
Platforms set their own rules, and some are stricter. Google’s Maps user-generated content policy prohibits offering incentives such as payment, discounts or free services in exchange for posting any review, and prohibits discouraging negative reviews or selectively asking only happy customers for reviews. So a gift card for a review, offered whatever the review says, may fall outside the FTC’s ban and still be against Google’s policy. Check the rules of each site you ask clients to use.
A simple way to stay well inside the lines.
- Ask every client at the same point, such as the end of a project, not only the ones you expect to be pleased.
- Offer nothing in return for the review.
- Never write a review for a client, even at their request. Offer a question or two to answer instead.
- Reply to every review, including critical ones, calmly and specifically.
- Do not ask staff, friends or family to post reviews. If a connected person does, the connection should be disclosed.
A short request template helps: “Thank you for working with us on [project]. If you have a few minutes, an honest review on [site] helps other people like you decide whether we are the right fit. Here is the link. Whatever you write, thank you.” It asks for honesty, names no rating, and goes to everyone.
What should you fix in the next 30 days?
One source at a time, starting with whichever a buyer sees first.
Work down the table in order. Each line names the source, what a referred buyer sees there, the usual fix, and how to confirm it worked.
| Source | What the buyer sees | Fix | How to check |
|---|---|---|---|
| Search for your name | Top results, a business profile if you have one, old listings | Claim your business profiles; correct or remove outdated listings you control | Search your name and “your name reviews” on a phone, signed out |
| Homepage | Whether you do what they were told you do | One plain sentence on what you do and for whom, above the fold | Ask someone outside the company to explain what you do after ten seconds |
| Service pages | Whether you do their specific thing | A page per service, in the words buyers use, with one example | Each service a referrer might mention has a page |
| About and team page | Who they would work with | Named leaders with current titles and short bios | Titles match LinkedIn exactly |
| LinkedIn company page | Whether you are active and current | Update the description to the fact sheet; post occasionally | The description matches your homepage sentence |
| Founder’s LinkedIn profile | Whether the person they were told about is credible | Current title, clear headline, a few recommendations | A stranger could tell what the company does from the profile |
| Review and directory sites | What clients say, and how recently | Ask every recent client the same way; reply to every review | New reviews appearing each month; none unanswered |
| Case studies | Proof for people like them | Two or three named examples, with permission | Each has a client, a problem, what you did and what changed |
| AI assistant, from memory | What it recalls about you, which may be old | Keep the sources above consistent; changes appear slowly | Run the five prompts monthly and note wrong facts |
| AI assistant, with web search | A summary of the pages it finds | Fix wrong facts at the page it cites; allow the search crawlers | Run the prompts with search on; check the cited pages |
Spread across a month, the work looks like this.
- Week 1: write the fact sheet. Run the search and AI checks and record what you find, as a baseline.
- Week 2: fix the website: homepage sentence, service pages, team page.
- Week 3: fix LinkedIn and every directory and review listing you control. Send the review request to recent clients.
- Week 4: rerun the checks, compare with week 1, and put the monthly check in the calendar.
How can the people who refer you help?
Your referrers are already doing your marketing. Make their job easier.
A referrer usually says something short and warm, and the buyer fills in the rest from the check. You can shape that first sentence. Give your best referrers a one-line description from your fact sheet and a link to the right page, so the name they pass on points somewhere that confirms it.
Clients who are happy to be named are the strongest proof you have. With their permission, a named case study on your site, a recommendation on the founder’s LinkedIn profile, or a mention on their own site gives a referred buyer, and an assistant searching the web, something concrete to find.
A referral hands you a buyer’s attention for a few minutes. Make sure everything they find in those minutes says the same true thing the referrer did.
See the use case for engineering and nearshore teams: How does a European nearshore team get chosen by Nordic and US companies?
Questions people ask next
My business grows through referrals. Why does my online presence matter?
Because referred buyers check you before they reply. They search your name, look at your website and LinkedIn, and sometimes ask an AI assistant. If what they find is out of date or inconsistent, the referral loses its force.
How do I find out what AI assistants say about my business?
Ask two or three assistants the questions a referred buyer would ask, with web search off and on, and write down the answers. Note wrong facts, missing proof and the pages cited. Repeat monthly, and fix errors at the source page rather than in the assistant.
Can I offer clients a reward for leaving a review?
Be careful. The FTC’s rule, in effect since 21 October 2024, bans incentives tied to a particular sentiment, positive or negative. Google’s Maps policy prohibits incentives for reviews altogether. The safest route is to ask every client the same way and offer nothing in return.
How long does it take for fixes to show up in AI answers?
Answers that use live web search can reflect a corrected page once it is crawled. Answers from a model’s memory change only when the model is retrained, which you do not control. Check both kinds monthly.
Can Throughline check what a referred buyer finds about us?
It can run the AI part of the check and draft the fixes. The Lab asks the questions a referred buyer would ask across the major assistants and Google’s AI Overviews, records what each says and which pages it cites, and Throughline drafts corrected page copy and review requests for you to approve.