Generative Engine Optimization: How to Get Your Business Cited by ChatGPT, Gemini and AI Search
- 18 minutes ago
- 7 min read
Something has changed in how people find businesses, and most companies have not noticed because their traffic numbers have not collapsed yet.
A growing share of research now happens inside an AI assistant. Someone asks ChatGPT which type of agency they need, or Gemini what a website should cost, or Perplexity to compare three platforms — and they get an answer with a handful of sources cited. They may never see a results page at all.
If your business is not in that answer, you are invisible for that query, regardless of where you rank in traditional search.
Generative Engine Optimization — GEO — is the work of being the source those systems cite. Here is what it actually involves, stripped of the hype that has accumulated around it in the last two years.
First, what has actually changed
Three things, and it is worth being precise because a lot of GEO advice is just SEO advice with a new label.
The answer is synthesised, not listed. Traditional search returns ten links and lets you choose. An AI assistant reads a set of sources and writes one answer. You are not competing for a position; you are competing to be material — the source that supplies a fact, a figure, or a framing that ends up in the answer.
Citation is the new click. Being named in an answer has value even when nobody clicks. A buyer who hears “studios like X specialise in this” has encountered your brand in a context of authority. That is closer to a recommendation than to a search result.
Queries are longer and more specific. People type keywords into search engines and sentences into assistants. “Web design agency Kolkata” becomes “I run a manufacturing business in eastern India and need a website that generates enquiries from distributors — what kind of agency should I look for?” Content built to answer specific, situational questions performs disproportionately well.
What actually earns citations
From what is observable across these systems, a consistent pattern emerges. Content gets cited when it is:
Specific and quantified. Assistants preferentially cite passages containing concrete figures, ranges and named specifics, because those are the parts that make an answer useful. “Websites in India cost between ₹30,000 and ₹6,00,000 depending on how much design and engineering is included” is citable. “Website costs vary depending on your requirements” is not.
Structured for extraction. Clear headings that pose real questions, direct answers immediately underneath, tables for comparisons, lists for sequences. If a passage can be lifted out and still make sense on its own, it is far more likely to be lifted.
Attributable to an entity the system recognises. These models weight sources by whether the publisher appears to be a real, consistent, identifiable organisation. Which brings us to the single most underrated part of this work.
Corroborated elsewhere. A claim that appears only on your own site is weaker than one that also appears in directories, industry publications, or third-party reviews. Consistency across the web is itself a trust signal.
Current. Assistants heavily favour recent content for anything time-sensitive, and are explicit about dates. An undated article is at a structural disadvantage.
The entity problem — the part most businesses skip
Before an AI system will cite you, it has to be confident you exist.
That sounds trivial. It is not. These models build an internal picture of your organisation from every mention of it across the web, and they discount sources where that picture is inconsistent. If your business appears as “Bulb Studio,” “Bulb Studios” and “BULB Design Studio” across your site, your LinkedIn, three directories and a press mention — with two phone numbers and two email domains — you are a fuzzy entity, and fuzzy entities get cited less.
The unglamorous fix, in order:
1. One canonical name, one address, one phone number, one primary email. Identical everywhere. Then actually go and correct the directories.
2. Organization schema on your site, with your legal name, logo, address, contact details and links to your social profiles. This is the machine-readable version of “here is who we are.”
3. A properly written About page that states plainly what you do, where, since when, and for whom. Assistants read these constantly.
4. Consistent presence in the places that are already trusted — Google Business Profile, LinkedIn, relevant industry directories, review platforms.
None of this is exciting and all of it is prerequisite. We put the site-level version of this in our B2B website checklist; for AI visibility it matters more, not less.
What to actually write
The content that earns citations looks different from content written to rank.
Answer the specific question in the first two sentences under the heading. Do not build up to it. If your H2 asks “How much does an app cost in India?”, the number should be in the next sentence. Assistants extract answers; they do not read your introduction politely.
Use real numbers and say where they came from. Ranges, percentages, timelines, thresholds. Sourced or from your own experience — both work, and both beat vagueness.
Write comparison content. “X vs Y vs Z” is enormously over-represented in AI citations, because comparison is exactly the task people bring to assistants. If you can compare things honestly — including saying when your option is the wrong one — that content will be cited.
Cover the questions nobody else answers. The obvious questions have a thousand answers, and you are one of a thousand candidates. The specific ones — “what does it cost to build an app in India versus the US,” “who owns the code when an agency builds your website” — have far fewer, and you can own them.
Date everything and update it. Publication and update dates, visible and in schema. For anything with a year in the title, put a genuine review cycle in the calendar.
The technical layer
Less than the SEO industry claims, but not nothing.
Structured data. Organization, Article, FAQPage, and Product or Service where relevant. This is unambiguous machine-readable meaning, and it is cheap to add.
Let the crawlers in. AI systems use their own crawlers, some of which are separate from Googlebot. If your robots.txt or your firewall blocks them, you have opted out. Check this — a surprising number of sites block AI crawlers by accident, and a few by a decision nobody revisited.
Do not hide content behind JavaScript. If the text only appears after scripts execute, some crawlers will not see it. Important content should be in the HTML.
Speed and accessibility still matter, for the ordinary reasons — and clean, semantic markup is easier for a machine to parse. There is more on the performance side in our Core Web Vitals guide.
On llms.txt: there is an emerging proposal for a file telling AI systems how to use your content. Support is inconsistent and it is not yet a reliable lever. Adding one costs nothing; expecting much from it would be premature.
How to tell whether any of this is working
Harder than SEO measurement, and worth setting expectations about.
Ask the assistants directly. Regularly put your key buying questions to ChatGPT, Gemini, Perplexity and Google’s AI Overviews and see who gets cited. Crude, manual, and by far the most informative thing you can do. Keep a simple record so you can see movement.
Watch for referral traffic from AI tools. It appears in analytics as its own referral sources. Volumes are typically small; quality is often high, because someone arriving from a synthesised answer has already done their research.
Watch branded search volume. If assistants are mentioning you, more people will search your name directly. This is often the earliest visible signal.
Track the questions, not the keywords. Build a list of the twenty questions a buyer asks before hiring someone like you, and check your presence in the answers to those. That list is a better strategy document than a keyword report.
The honest assessment
Much of GEO is good SEO practised with more discipline: be genuinely useful, be specific, be structured, be consistent about who you are, be current.
But two things are genuinely new, and worth acting on.
The first is that being a coherent, verifiable entity now has direct commercial value. Sloppy consistency across your web presence used to be a minor untidiness. It now affects whether a machine is confident enough to name you.
The second is that the reward for specificity has gone up sharply. Generic content had residual value in traditional search — it could still rank on domain authority. In a synthesised answer there is no long tail of positions. Either you contributed something to the answer or you did not.
Which is, in a roundabout way, encouraging. The strategy that works is to know something worth citing and to write it down clearly. That has always been the advice. It is just now enforced.
Frequently asked questions
What is generative engine optimization? The practice of making your content likely to be used and cited by AI systems such as ChatGPT, Gemini, Perplexity and Google’s AI Overviews when they answer questions in your area. It overlaps heavily with SEO but optimises for being a source rather than for a ranking position.
Is GEO different from SEO? It shares most of its foundations — useful content, clean technical structure, authority. The differences are an emphasis on extractable, specific, well-structured passages, and on being a consistent, recognisable entity across the web rather than just a well-linked domain.
How do I get my business mentioned by ChatGPT? Publish specific, quantified, well-structured content answering the questions your buyers actually ask; make sure your business details are identical everywhere online; add Organization and FAQ schema; and make sure AI crawlers are not blocked. Then track which questions you appear in and write toward the gaps.
Does GEO replace traditional SEO? No. Most search still happens conventionally, and the two overlap substantially. Treat GEO as an additional lens on the same work rather than a separate programme.
How long does it take to see results from GEO? AI systems update on varying schedules, so weeks to months rather than days. Entity consistency fixes tend to show up soonest; content-driven citation gains take longer.
Should I add an llms.txt file? It costs nothing and may help as support grows, but adoption is inconsistent and it is not currently a reliable lever. Fix your entity consistency and content structure first.
Curious whether AI assistants can even read your site properly? Our free website audit now covers AI-search readiness alongside conversion and performance. You’ll get a Conversion Leak Report with the specifics for your site. Request your audit →