The short version: ChatGPT, Perplexity, and Google’s AI Overviews are already answering questions about your business, your prices, and your competitors, whether you’ve done anything about it or not. Most small business owners have never checked what these tools are saying, and when they do, they often find mistakes that have been quietly costing them enquiries for months.
The email that started this whole thing
In November a woman emailed me asking about my day rate. Nothing unusual there, except she’d already made her mind up before she wrote to me. “I saw you charge from £3,000,” she said. I don’t. I never have. When I asked where she’d seen that, she said she’d asked ChatGPT to compare a few UK marketing consultants and it had given her a figure, a bio, and a list of clients, some of which weren’t mine.
I went and checked. It wasn’t malicious, and it wasn’t even that far wrong, but it had pulled together bits of old bios, a press mention from 2019, and something from a directory site I hadn’t updated in years, and stitched them into a confident, wrong answer. No hedging, no “I’m not sure”, just a clean, plausible paragraph that happened to be inaccurate in three places.
That’s when it clicked for me that I’d been treating AI visibility as a future problem to think about eventually, when it was already an active, present problem with a real cost attached to it. If someone had decided not to email me because of a number that isn’t even true, how many other people had quietly done the same thing without telling me?
This is not the same problem as SEO, and treating it like SEO wastes your time
A lot of what gets written about “AI SEO” is just old SEO advice with the word AI bolted on. It isn’t the same game. Google Search still mostly ranks pages. AI answer engines summarise across dozens of sources at once and often prefer places you don’t control at all.
When I ran a handful of “best [industry] near me” and “best [service] for small business” style prompts through ChatGPT and Perplexity for six of my own clients in January 2026, the sources being cited most often weren’t company websites. They were Reddit threads, review aggregator pages, LinkedIn posts, and Wikipedia where it existed. Company homepages showed up, but usually third or fourth, and often with facts that were two or three years out of date.
That’s the uncomfortable bit nobody selling “AI visibility packages” wants to say out loud: you can tidy up your own website endlessly and it will only move the needle a little, because the AI models are frequently pulling their confidence from places you don’t own and can’t directly edit. You can influence those places. You can’t control them.
What’s within your control
There’s still plenty you can do, and most businesses have done none of it. Here’s the order I work through with clients now, roughly in the sequence that gives the best return for the least effort.
1. Check what’s already being said, before you fix anything
Open ChatGPT, Perplexity, and Google (with AI Overviews) in separate tabs. Ask each one the questions a prospect would type:
- “What does [your business name] do and who are they best for?”
- “How much does [your service] cost from [your business]?”
- “Who are the best [your category] in [your city]?”
Write down every factual claim in the answers. Prices, locations, years in business, client names, awards. You’re looking for anything wrong, outdated, or attributed to a competitor by mistake. I do this for every new client now before we touch a single page of their site, because there’s no point optimising content if the AI is confidently quoting a 2022 price list from a directory that hasn’t updated since.
2. Find the source of the wrong information and fix it there
When an answer is wrong, it usually traces back to one of a small number of places: an old press release, a directory listing (Clutch, Yell, Bing Places, a local chamber of commerce site), a cached version of an old page, or a review site with an outdated business description. Fix it at the source, not just on your own site. I found the £3,000 figure was sitting on a “top consultants” directory page that hadn’t been touched since 2021. One email to the site owner and a corrected listing, and that specific error dropped out of AI answers within about three weeks.
3. Make your own facts boringly, repetitively consistent
AI models weight information more heavily when the same fact appears identically across multiple places. If your website says “since 2015,” your LinkedIn says “founded 2016,” and your Companies House filing says something else again, you’re handing the model a coin to flip. Pick the exact wording for your founding date, your pricing structure, your location, and your specialism, and use that exact phrasing everywhere: your site, your Google Business Profile, LinkedIn, any directory listing you can edit. This is the same discipline behind brands whose facts never seem to get garbled online. Look at how consistently Mailchimp built its brand around one clear positioning for small business email marketing, repeated the same way for years, so there’s very little for any tool, human or AI, to get confused about.
4. Give the models something specific to quote
Vague marketing copy gives AI nothing usable to summarise, so it goes looking elsewhere. Specific numbers and named details get lifted and repeated. “We help small businesses grow” produces nothing. “We work with businesses turning over £200,000 to £2 million who want to add AI to their marketing without hiring a full team” gives a model an actual sentence to quote back. This is the same instinct behind good FAQ content and interactive tools, the kind of specific, self-contained answer people build things like interactive calculators around, because a concrete, quotable answer travels further than a paragraph of soft positioning ever will.
5. Get mentioned on the sites AI trusts
Since Reddit, review sites, and industry roundups outrank most company websites in AI answers, the highest-use work often isn’t on your own site at all. Answer questions on relevant subreddits and forums using your real name and business (not spammy self-promotion, actual useful answers). Ask happy clients to leave detailed reviews on Google and Trustpilot that mention specifics, not just “great service.” Get quoted in roundup articles and local press. This is digital PR wearing a new hat, and it’s exactly the kind of consistent presence-building that content marketing people like Joe Pulizzi were telling businesses to do years before anyone was worrying about AI answer engines. Becoming the obvious, most-referenced source in your niche was always good advice. It’s just got a new, more urgent reason to matter.
6. Keep a Wikipedia-adjacent presence honest, even if you’re too small for Wikipedia itself
Most small businesses will never qualify for a Wikipedia page, and that’s fine. But Crunchbase, LinkedIn company pages, and industry-specific databases function the same way for AI models, as a trusted reference point. Fill these out, keep them current, and treat outdated entries as urgently as you’d treat a wrong price on your homepage.
7. Set a recurring check, not a one-off audit
This isn’t a project with an end date. I now check my own AI answers monthly, on the same day I check my analytics. It takes fifteen minutes and it’s caught two more small errors since November. Put it on the calendar or it won’t happen.
The bit that stings a little
Here’s the part most people don’t want to hear: fixing all of the above will not save a weak offer, a confusing website, or prices nobody wants to pay. I’ve seen business owners spend weeks chasing AI mentions while their actual conversion rate sat at under 1%. Getting cited accurately by ChatGPT gets you in front of more of the right people. It does nothing for you if what they find once they click through is a slow, cluttered site with no clear next step. Fix the fundamentals first. AI visibility is an amplifier, not a fix.
It’s also worth saying plainly that some of this is out of your hands. You can’t force Reddit to talk about you. You can’t make a journalist quote you accurately. You’re influencing probabilities, not flipping a switch, and anyone promising guaranteed AI rankings is selling you the same nonsense that used to get sold as “guaranteed page one Google rankings” a decade ago.
Where the community angle helps
One thing that consistently moves the needle faster than anything else on this list is an active, visible community around your brand, because AI models pick up patterns of repeated, organic mentions much faster than they pick up a single polished page. Look at how GoPro built its marketing almost entirely on customer-generated content, or how Figma grew through its design community rather than paid ads. Both brands have thousands of independent voices describing them the same consistent way across forums, YouTube, and social platforms. That’s a wall of consistent third-party signal that’s very hard for an AI model to get wrong, and it’s the same reason a well-run Instagram marketing strategy with a consistent bio, pinned facts, and repeated messaging quietly helps your AI visibility even though that was never the point of the account.
When it’s worth getting help instead of doing this yourself
If you’ve got one location, one clear service, and twenty minutes a month, you can run this whole process yourself with the steps above. Where it gets harder is multi-location businesses, agencies with several service lines, or anyone who’s had three different brand names or a merger in the last few years, because the conflicting-facts problem multiplies fast. That’s usually when I get called in, and it’s also the point where working with an AI consultant for small business earns its fee, because untangling years of inconsistent listings across forty directories is not a fifteen-minute job, it’s a proper clean-up project with a checklist and a deadline.
Frequently asked questions
How do I know if ChatGPT is even talking about my business at all?
Ask it directly, using the exact phrases a customer would type, not your own brand name in isolation. “What does [business name] do” often gives a different, more revealing answer than “Tell me about [business name].” If it gives you nothing specific or hedges heavily, that’s useful information too, it means you’ve got a blank slate to fill rather than misinformation to correct.
Can I fix a wrong answer in ChatGPT directly?
Not directly and not instantly. You fix the sources it’s pulling from (directories, old press mentions, review sites) and update your own site and profiles with consistent facts, then wait for the models to refresh their training or retrieval data. In my experience that’s taken anywhere from two to six weeks for a correction to stop appearing in answers, depending on the tool and how many places the wrong fact was repeated.
Does this replace normal SEO or do I need to do both?
Both, and they overlap more than they conflict. Good, specific, well-structured content that ranks on Google also tends to get pulled into AI answers, because a lot of these tools are still using search results as part of their sourcing. Don’t abandon SEO for “AI SEO.” Do the same solid work, but also check the AI tools separately because they weight sources differently, and sometimes prefer Reddit or a review site your normal SEO checklist never touches.
Is it worth paying someone to manage this, or can a small business do it alone?
A single-location business with one service line can run the monthly check and fix outdated directory listings without help. It’s mostly patience and attention to detail. Where it’s worth paying for support is when you’ve got multiple service lines, multiple locations, or years of inconsistent branding to untangle, because that’s a genuine clean-up project, not a fifteen-minute monthly task.
Related reading: Why AI Pricing Advice Is Quietly Training Small Business Owners to Undercharge and How to Reverse-Engineer Any Competitor’s Marketing Strategy Using AI (Without Paying for Fancy Tools).
