How to get your business cited by ChatGPT, Perplexity and Google AI Overviews (GEO explained)

AI answer engines cite businesses whose pages they can crawl, whose facts they can lift in a sentence or two, and whose claims are backed up elsewhere on the web. Generative engine optimisation (GEO) is the work of making all three true. Most of it builds on solid SEO; the rest is about writing answers rather than pages.

What GEO is, and how it differs from SEO

Search engine optimisation earns you a position on a results page. Generative engine optimisation earns you a mention or a citation inside an answer that a large language model has written for the user. That user may never see a list of ten blue links at all.

The two disciplines overlap heavily. Google AI Overviews draw on the same index as regular search, Perplexity runs its own live web retrieval, and ChatGPT with search combines what the model learned in training with fresh results from its own crawler. If your site is slow, blocked or thin, none of them will find you. But ranking well is not enough on its own. An answer engine reads a page looking for a passage it can quote, verify and attribute. Pages built purely to rank often do not contain one.

How the three engines choose sources

Each platform works differently, and the differences shape where you spend effort.

• Google AI Overviews lean on pages that already rank near the top for the query, then favour those with clear passage-level answers and structured data. If you are not on page one, you are rarely in the overview.
• Perplexity retrieves fresh pages for every query and shows numbered citations. It rewards recency, specificity and pages that answer the exact question asked, and it is noticeably more open to smaller domains than Google.
• ChatGPT blends training knowledge with live search. Brand mentions across the wider web, consistent business facts and third-party references decide whether it names you at all.

The practical upshot: Google rewards authority you already have, Perplexity rewards a well-written answer today, and ChatGPT rewards a consistent footprint across many sites over time.

Step one: let the crawlers in

Before touching content, check access. A surprising number of Canadian business sites block AI crawlers by accident, through a firewall default or a robots.txt rule copied from a template.

• Confirm that GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot and Google-Extended are not disallowed in robots.txt, unless you have chosen to block them deliberately.
• Check that your CDN or security layer is not serving challenge pages to those user agents.
• Make sure the important content is in the HTML on first load. Text that only appears after JavaScript runs is often skipped.
• Publish an llms.txt file at the root of the domain listing your most important pages with a one-paragraph description of the business.

Step two: write answers, not just pages

Answer engines extract passages. The pages cited most often share a shape: a direct answer in the first two or three sentences, then the reasoning, the specifics and the exceptions. That is the opposite of the classic marketing page, which builds slowly towards its point.

Lead with the answer

Every service page and every article should open by answering the question a searcher would type. A page about web app pricing should not begin with a paragraph on your values. It should explain how a build is scoped, what drives the effort and how to request a quote.

Use question-shaped headings

Headings that mirror real queries (“How long does an iOS app take to build?”) help both the model and the reader find the passage. Keep each section self-contained so it still makes sense when quoted on its own.

Be specific and verifiable

Models prefer claims they can cross-check. “Founded in Vancouver in 2014” is citable. “Trusted by businesses everywhere” is not. Include dates, locations, the standards you build to (WCAG 2.2 AA, for example) and named methods rather than adjectives.

Add FAQs that deserve to exist

A short FAQ block at the foot of a page, marked up with FAQPage schema, hands engines a set of clean question-and-answer pairs. Write the questions people actually ask, including the awkward ones about timelines and cost.

Step three: make your facts consistent everywhere

Language models assemble a picture of your business from many sources at once. If your address, founding year, services or even your name vary between your website, your Google Business Profile, LinkedIn, directories and press coverage, the model treats you as uncertain and moves on to a competitor it is sure about.

• Use one canonical business name and description, and repeat it word for word across every profile.
• Publish Organization, LocalBusiness and Service schema with the same values as the visible page.
• Keep an About page that states plainly what you do, where you are, since when and for whom.
• If you serve both official languages, publish parallel English and French pages with hreflang so each engine can serve the right one.

Step four: earn mentions off your own site

Citations are partly a popularity contest, but the currency is mentions rather than links. Being named in an industry association’s member directory, a local business journal, a conference speaker list or a well-read comparison article all count, whether or not a link is attached. Detailed reviews help too. Ask clients to describe what you built for them, not just how pleasant you were to work with.

Step five: measure it

Build a list of 30 to 50 questions a prospective client might put to an AI assistant about your category (“Who builds bilingual member portals in British Columbia?”), then run them monthly across ChatGPT, Perplexity and Google, logging whether you are mentioned, cited or absent. Track referral sessions from those platforms in your analytics as well. Both numbers move slowly at first, then quickly.

What a GEO engagement looks like

When we take this on for a client, it runs like every other project at OlDevs: a strategy call to identify the questions that matter commercially, an audit of crawler access, schema and content, then a build phase where we rewrite priority pages and ship them with weekly working demos so you see progress rather than a slide deck. A regional professional association we worked with saw its AI citation rate for member-facing questions go from near zero to a consistent presence within one quarter. As always, you own every page, every schema file and every account we touch.

GEO is not a trick and it is not a separate website. It is the discipline of being the clearest, most verifiable source for the questions your customers ask, and then making sure every machine that reads the web can see it. Start with access, then answers, then consistency. The citations follow.