Getting cited by ChatGPT, Perplexity, Claude, or Google's AI Overviews comes down to four practical things: write passages an AI can lift verbatim, publish structured data the AI can parse, get mentioned honestly in places AI models read, and keep your traditional SEO clean. None of these require a specialist. All of them require discipline. This is the field guide.
A note on what NOT to do first: don't pay for "GEO services" until you've shipped the four things below. Almost everything the consulting industry will charge for is one of these four with a markup.
One: write passages an AI can lift verbatim
AI models cite passages, not pages. The unit of citation is a self-contained chunk of text — usually 1-3 sentences — that fully answers a question on its own. If your page buries the answer in paragraph five, the model is going to lift the answer from somewhere else.
The format that works:
- H2 heading is a real question someone would ask ("How long does small business SEO take to work?")
- First sentence after the heading is a literal, scannable answer with a number ("Three to six months before meaningful movement, twelve months for compounding work to pay back.")
- Following paragraphs add nuance, but the lead sentence stands alone
Look at your existing service pages and check whether the first sentence of each section is liftable. Most aren't. Fix the leads and you've done more for AI visibility than most agencies do in a quarter.
Two: publish structured data the AI can parse
Schema markup is what tells a machine "this is a Service costing $99/month, here are the FAQs, here is the area served." Without schema, the AI has to guess from prose. With schema, the AI knows.
The four schema types that matter most for a small business:
- Service with
offerscontaining explicitpriceandpriceCurrency - FAQPage with
mainEntitycontaining each Q&A - BreadcrumbList on every internal page
- Organization on the site root, with
sameAspointing to your LinkedIn, your Maps profile, and any other canonical brand profiles
You can validate your existing schema with Google's Rich Results Test in about three minutes. If you have zero schema or broken schema, that's your single highest-leverage AI-visibility move this month.
Three: get mentioned honestly in places AI models read
AI models train on the open web. Reddit, Hacker News, Indie Hackers, niche industry forums, Quora — these are the surfaces that disproportionately shape what an AI model believes about your industry. A brand mention in a Reddit thread about "good agencies that don't charge monthly" is worth more for AI citation than a paid link from a directory site.
The discipline: spend ten minutes a week being genuinely useful in two or three threads where your industry lives. Don't astroturf. Don't drop your URL in every comment. Just answer questions, share your honest experience, and let the brand mention happen naturally when it's relevant. Six months of this changes what AI models say about you.
Mule's own AI visibility is downstream of being mentioned in threads about small-business websites, owner-of-record, and rural digital agencies. None of those mentions were paid for. All of them happened because the team is honest about what we do and don't do.
Four: keep traditional SEO clean
Most GEO problems are SEO problems. Pages that load slowly are cited less. Pages with broken schema are cited less. Pages with low E-E-A-T signals (no author, no date, no expertise markers) are cited less. Pages that don't appear in Google at all are not cited because the AI didn't crawl them.
So before you optimise for the AI search surface specifically, make sure the basics work: pages indexed in Google Search Console, no broken links, Core Web Vitals in the green, author bylines on every blog post linking to a real person, organization schema with sameAs pointing to verified profiles. The boring stuff. AI models reward it because their training data does.
What changes month over month
You won't see AI citation visibility ramp the way Google rankings do. AI models are batch-trained, so changes you make today might not show up in the model's responses for weeks or months. The honest expectation: six months of consistent work before you start seeing your brand named in ChatGPT or Perplexity answers about your category.
The companion glossary entries at /glossary cover GEO and AEO definitions. The broader explainer about how GEO works for small businesses is at /blog/generative-engine-optimization-small-business. If you want the kind of ongoing technical work that includes AI-search optimisation in a real $99/month retainer, the page at /cheap-seo-services walks through what's actually in scope.