To get cited by ChatGPT, Perplexity, and Google's AI Overviews, do four things: mark your pages up with schema.org JSON-LD, keep your business name, address, and description identical everywhere they appear, answer each question in the first two sentences under its heading, and earn third-party mentions that corroborate what you claim.
The awkward part is that you can't "rank" in ChatGPT. There's no page two to climb off of. The model either cites you, mentions you by name, or has no idea you exist. The practice of tilting those odds goes by GEO (generative engine optimization) or AEO (answer engine optimization), and I'll be upfront: it's young, and anyone claiming a reliable playbook is ahead of the evidence. Nearly all of what does work builds on ordinary SEO rather than replacing it — the same fundamentals covered in our guide to SEO for small businesses in 2026, pointed at a newer set of readers.
The channel is worth the attention. Adobe Analytics measured traffic to US retail sites from generative AI sources during the 2024 holiday season and found it up more than 1,300% year over year. A number like that from a small base still deserves skepticism (my houseplant collection also grew 300% this year; I own four houseplants), but it describes a real thing: an acquisition channel forming while most businesses aren't looking.
How do AI search engines choose which sources to cite?
A language model answers questions from two sources: what it absorbed during training, and what it retrieves live from the web when a query needs current information. Retrieval is where your website gets its shot.
When ChatGPT runs a live search, or Perplexity assembles an answer, the system pulls indexed pages, weighs relevance, and picks sources. The behavior is less like classic keyword ranking and more like a fast research assistant skimming a stack of open tabs. Pages that get cited tend to share traits: easy to skim, unambiguous about who's behind them, structured so they're hard to misread.
The encouraging version of this: a site that's genuinely useful to a human skimmer is most of the way to being useful to a model. The rest is specifics.
One specific worth naming early: retrieval bots are impatient, and a page whose content only appears after several seconds of JavaScript may be read as thin or skipped entirely. The same work that fixes Core Web Vitals like LCP, INP, and CLS — server-rendered content, fast first paint, no layout shifting — makes your pages easier for a crawler to read in one pass.
Structured data: label what the machine is looking at
Schema.org markup as JSON-LD has been standard good practice for years, and the AI-search shift makes it more valuable, not less.
Every crawler — search engine or AI — is constantly asking one question about your pages: what kind of thing is this? Without structured data it has to infer. With it, you've answered directly.
For most small businesses, the markup worth having:
- Organization or LocalBusiness — legal name, address, phone, hours, logo, consistent on every page
- FAQPage on anything that answers common customer questions
- Product for e-commerce, with price, availability, and reviews
- Article on blog content, with author and publish date
None of this is exotic, and Google's Rich Results Test or Schema.org's validator will tell you within minutes whether it's implemented correctly. If nobody has touched your structured data since launch, an audit is overdue — it's core to how we approach SEO on every build.
Entity clarity: be one unambiguous thing
Search engines and AI systems both try to resolve your business into an "entity" — a single identifiable thing with a consistent name, location, and description across the web.
Every inconsistency fights that resolution. A business name written five slightly different ways across your site, Google Business Profile, Yelp, and Facebook. Two old phone numbers still floating around directories. An about page that describes what you did in 2019.
The fixes are boring and effective:
- 1.Pick one canonical business name. Use it everywhere — site, directories, social profiles, email signatures.
- 2.Write a plain "About" page. Not marketing copy: who you are, what you do, where you're located, who you serve. When someone asks a model "who is [your business]?", this page is what it reaches for.
- 3.Make your name, address, and phone identical across your site, Google Business Profile, and the major directories. Local SEO has demanded this for a decade; AI retrieval wants the same thing.
- 4.If you're a legitimately recognizable authority in your space, a Wikipedia or Wikidata entry — where actually warranted — carries unusual weight, because both are heavily used in training and retrieval.
Does llms.txt actually do anything?
You've maybe seen llms.txt pitched as "robots.txt for AI" — a plain-text file at your site root summarizing what the site is and linking your key pages in model-friendly markdown.
The current reality: it's a proposed convention, not a standard. OpenAI, Anthropic, and Google have not committed to reading it the way search engines universally honor robots.txt and sitemap.xml. Some crawlers check for it. Many don't.
I'd still write one. It takes an hour — short site description, links to services, about, contact, top articles — and if adoption grows, you're already in place. If it doesn't, you lost an hour (roughly what you'd have spent refreshing your rankings anyway). What you shouldn't do is treat it as the strategy. The file is a pointer; it can't compensate for the pages it points to being thin.
Answer the question where the reader lands
This is the piece that matters most and gets the least attention, probably because it's writing work rather than a technical checkbox.
Answer engines favor content that states the answer near the top. Someone asks "how much does a website redesign cost" — a page that opens with a real range and then explains the variables gets pulled into answers. A page that opens with three paragraphs on the history of web design does not.
If this sounds like featured-snippet advice from five years ago, it basically is. AEO just raises the stakes on it. Concretely:
- The first sentence or two of any section should stand alone as an answer.
- Real customer questions become headings, phrased the way customers phrase them.
- State numbers, prices, and specifics directly. "Contact us for a quote" where a range would be honest is a citation you handed to a competitor.
- One clear topic per page. A page straddling five services is hard for anything — human, crawler, or model — to categorize with confidence.
A cheap diagnostic: put your own likely customer questions into ChatGPT and Perplexity and see who gets cited. (Budget an hour. This exercise has a way of becoming an afternoon.) If the answer is consistently "not you," now you know, and you know who to study.
Third-party corroboration still decides trust
Models, like people, weight independent evidence. A business that exists only on its own website reads differently than one with Google reviews, local press mentions, industry directory listings, and a footprint on relevant third-party sites.
This is the same groundwork as traditional local SEO — citations, reviews, backlinks — and it didn't stop mattering when AI search arrived. If anything the weighting went up, because a model deciding whether to cite you by name is looking for exactly this kind of corroboration.
What should you avoid when optimizing for AI search?
Don't stuff pages with keyword variations aimed at gaming an AI crawler. Language models are notably good at recognizing padded, unnatural text, and the humans who still make up the overwhelming majority of your traffic have to read it too.
Don't buy from anyone promising guaranteed ChatGPT placement. Nobody outside these companies controls what a closed model cites (judging by their published research, control is loose even inside them), and certainty is the tell that you're being sold. The fundamentals — structured data, entity consistency, direct answers, real corroboration — shift odds. That's all anything does right now.
And don't reallocate your whole SEO effort here. Google still drives the large majority of commercial search traffic in 2026. AI search is growing fast from a small base — worth building toward, not worth abandoning the main channel for.
This week's version
Three moves, in order: audit your structured data, tighten entity consistency everywhere your business appears, and rewrite your top three service pages so the first two sentences under each heading actually answer the heading. Our free tools can flag the quick technical wins on your current setup.
None of it is a silver bullet. It's the discipline that has always driven good SEO — clarity, consistency, content worth citing — aimed at a newer set of readers. Get that right and you're positioned for whichever lane keeps growing.
Want your site built to be found by Google and by AI search? Structured data, clean entity signals, and answer-first content are baked into every site we build. See our SEO services or get in touch to talk through where your site stands today.