Straight answer: ChatGPT and Perplexity mostly ignore small business content because it reads like everyone else's content, not because of some hidden technical setting. Fix that by putting real numbers, named examples, and a direct answer in the first two sentences of every page. And go in knowing that even when you do get cited, the click-through back to your site will likely stay small for years yet.
What "being cited by AI" means for a small business
When someone asks ChatGPT "who's a good AI consultant near Reading" or "what does a kitchen refit cost in 2026", the model pulls from pages it has crawled and, increasingly, from live search results it checks in real time. If your page turns up in that answer, with your business name attached, that's a citation. No click required for it to count as a mention.
Business owners hear this and panic-buy "AI SEO" services that mostly repackage old technical SEO advice with a new label. I get asked about this constantly by clients who've read a LinkedIn post claiming they need schema markup on every page or their business "won't exist" to AI. Schema helps. It is not the reason most small business content gets skipped.
The kitchen fitter in Tunbridge Wells
A client of mine, a kitchen and bathroom fitting business, had 40-odd blog posts. Standard stuff: "5 Signs You Need a New Kitchen", "Benefits of Quartz Worktops", the usual. I ran a batch of the exact questions their customers ask through ChatGPT and Perplexity: "how much does a kitchen refit cost in Kent", "how long does a bathroom renovation take", "best worktop for a small kitchen". Not one of their 40 posts came up. Competitors with worse writing but real prices did.
We rewrote six posts. Not all 40, just the six that matched real search questions we could see in their Google Search Console data. Each one opened with a direct answer in the first line: "A full kitchen refit in Kent typically costs between £8,000 and £22,000 depending on layout changes and worktop choice." Then the detail underneath. Within about ten weeks, Perplexity started citing two of those posts by name when I re-ran the test queries. Their site traffic from AI referrers went from zero to roughly 15 to 20 sessions a month. Not thousands. Twenty.
That number matters, because it's the uncomfortable bit nobody selling "AI visibility packages" wants to say out loud.
The referral traffic is small, and it will stay small for a while
Here's the truth most people flogging AI SEO courses skip past: getting cited by ChatGPT or Perplexity does not send you meaningful traffic yet, for most small local or regional businesses. Google still sends the volume. AI assistants send trickles, usually in the tens or low hundreds of monthly sessions even for sites doing this well, because a huge share of AI answers satisfy the question on the spot and the user never clicks through at all. That's the whole point of the tool from the user's side.
So why bother? Two reasons. First, being the named source inside an AI answer is a trust signal even without a click, the same way being quoted in a newspaper builds credibility whether or not anyone visits your site that day. Second, this is early. The businesses building citable, specific content now are the ones who'll be sitting in the answer boxes when volume does grow, the way early bloggers who wrote real how-to content in 2010 are still ranking for it now.
I'd rather tell a client "this will get you 20 extra visits a month for now, and a stronger position later" than sell them a fantasy. That's the version of this that's true, and it's worth doing anyway.
Why generic content gets skipped
Large language models are built to summarise and synthesise. When ten sites say the same vague thing, "quartz worktops are durable and low maintenance," the model has no reason to pick you over the other nine. It'll either paraphrase the general consensus with no attribution or grab whichever source has the clearest, most specific claim.
Specific beats polished, every time. Compare how a proper marketing case study reads. My breakdown of the GoPro marketing strategy works because it names dates, campaign names, and real growth figures rather than saying "GoPro built a strong brand." Same with the Figma marketing strategy piece, or the one on Mailchimp's marketing strategy. A model can lift a fact from those posts and attribute it, because there's an actual fact to lift. "They grew fast through good marketing" gives it nothing to quote.
The Instagram marketing strategy post is a good test case too. It survives because it names specific product launches and specific dates rather than describing Instagram in general terms. That's the pattern: named, dated, numbered.
Step by step: auditing whether AI can even find you
Before touching a word of content, check whether the crawlers are reaching your site at all.
- Check your server logs or ask your host to pull them, and look for GPTBot, PerplexityBot, and ClaudeBot hits. If none of the three have visited in the last 30 days, your content isn't in the running regardless of quality.
- Check your robots.txt file. A shocking number of small business sites, especially ones built by an agency years ago, have blanket disallow rules that block these crawlers by accident.
- Pick 10 real questions your customers ask, the actual phrases, not keyword-tool guesses, and run them through ChatGPT, Perplexity, and Google's AI Overview. Note who gets cited.
- Open your three highest-traffic blog posts and time yourself: how long until a specific number, name, or fact appears? If it's more than two sentences, that's your fix.
- Rewrite the opening of each post as a direct, quotable answer to the exact question you're targeting. One sentence. Then expand underneath with the how and why.
That fifth step is the one people skip because it feels too simple to matter. It's the one that moved the needle for the kitchen fitter.
Build something the AI has to point to you for
The strongest move isn't better writing, it's owning a piece of data nobody else has. An interactive tool that generates a personalised number, a cost calculator, a quiz that gives a real result, creates something a language model can describe but not replicate, which pushes it toward linking back to you as the source rather than just summarising a paragraph. I've written before about how interactive calculators pull in and convert visitors, and the same logic applies here: a calculator that spits out "your estimated kitchen refit cost is £X" based on inputs is a genuine asset that both humans and AI systems treat as a primary source, not a rehash.
Original survey data does the same job. Survey 50 of your customers, publish the actual numbers with your methodology stated plainly, and you've created a fact that didn't exist before you published it. Models have nowhere else to pull that fact from except your page.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The part about content strategy nobody wants to hear
This is where I'll say the thing that annoys people who sell content-at-scale packages: publishing more posts is not the fix. A site with 12 sharp, specific, number-heavy posts will out-cite a site with 200 generic ones nearly every time, because volume was never the signal, specificity was. Joe Pulizzi built an entire content marketing philosophy around this exact idea long before anyone was talking about AI citations, and the lessons from his career hold up better now than when he first said them: pick a narrow enough topic that you can be the best answer for it, not one of a thousand adequate ones.
If you're paying someone £500 a month for "four AI-optimised blog posts," ask them one question: which real number, name, or dataset is in each post that isn't in a competitor's post. If they can't answer that in a sentence, you're paying for volume that won't get cited.
When it's worth getting help
If you've run the audit above and found the crawlers reaching your site but never citing you, that's usually a content specificity problem, fixable in-house with the steps I've laid out. If you've found the crawlers aren't reaching you at all, or you don't know where to start pulling server logs, that's a technical and strategic job worth bringing someone in for. I work through exactly this with clients as part of AI consultant work for small businesses, usually starting with the same 10-question audit described above before we touch a single page.
What to do this week
- Pull your server logs and confirm GPTBot, PerplexityBot, and ClaudeBot are visiting.
- Run your top 10 customer questions through ChatGPT and Perplexity and write down who gets cited.
- Rewrite the opening line of your three best-performing posts as a direct, numbered answer.
- Pick one thing you can turn into original data: a calculator, a small customer survey, a real price range from your last 20 jobs.
- Set a reminder to re-run the same 10 questions in eight weeks and compare.
Frequently asked questions
Does getting cited by ChatGPT bring in customers?
Not many yet, directly. Referral traffic from AI assistants to small business sites is usually small, often under a few dozen sessions a month even when citation is working well, because most people get their answer without clicking through. The value right now is trust and early positioning, not a traffic surge.
Do I need schema markup to get cited by AI?
It helps but it isn't the deciding factor. Content that gives a direct, specific, numbered answer to a real question gets cited whether or not schema is present. Schema is a small boost on top of good content, not a substitute for it.
How is this different from normal SEO?
Normal SEO optimises for ranking in a list of blue links a person scans. AI citation optimises for being the single fact or example a model chooses to quote inside a written answer. Both reward specificity, but AI citation punishes vagueness much harder because there's no second chance further down the page.
How long does it take to see AI referral traffic change?
In client work, meaningful shifts in citation have shown up within 8 to 12 weeks of rewriting a handful of targeted pages, provided the crawlers were already reaching the site. If crawlers weren't reaching the site at all, fix that first, the content changes won't matter until they can.
Related reading: Why Every AI-Written Newsletter Sounds the Same (And What I Changed In Mine) and The AI Sameness Problem: Why Your Small Business Marketing Sounds Like Everyone Else's.
I go much deeper on this in the AI marketing guide.
Before you change anything, test your AI visibility so you have a baseline to measure against rather than a guess.
