Here's a question nobody asks until it's too late: when your ideal buyer opens ChatGPT and types "best SEO agency in Montreal," whose site does the answer name? If the answer is not you, you're losing a conversation you didn't even know you were in.
And here's the uncomfortable part. ChatGPT search doesn't work like Google. There are no positions one through ten waiting to be climbed. There's just an answer and a set of citations underneath it. You win by being the source the answer is built from, not by landing somewhere in a ranked list.
How ChatGPT selects sources
When ChatGPT searches, it gathers content and decides what to quote. The citations under the answer are chosen in part by the same index-level signals that power traditional search, but the choice of what to say is driven by how cleanly your content can be extracted.
That's why we see sites with mediocre rankings getting cited while heavy-hitters get skipped. The cited site stated the answer plainly. The heavyweight buried it.
ChatGPT's source selection process considers:
- Content clarity — Can the AI extract a direct, quotable answer?
- Entity recognition — Can ChatGPT confidently identify your brand as a distinct entity?
- Cross-web consistency — Do multiple sources agree about who you are?
- Authority signals — Do other sources reference and cite your content?
- Structured data — Does your site provide machine-readable context?
The key insight: ChatGPT doesn't just evaluate your page in isolation. It cross-references your website with other sources it can find. If your website says one thing and your Google Business Profile says another, the uncertainty reduces your chances of being cited. Consistency across the web is the foundation that makes everything else work.
For a deeper look at how different AI engines select sources, see our guide on what AI visibility is and how to measure it.
Understand what the "rankings" actually are
Getting extracted is the threshold our AI Visibility Alerts measures, and these are the three things that control it:
Three things control whether you get extracted:
Answer in the open. The response to the question should appear near the top of your page, in a plain sentence, not behind a vague lead-in. AI engines are bad at inferring; they're very good at cutting a clean sentence and running.
If someone asks "how much does SEO cost in Canada," and your pricing page buries the answer below a hero image and three paragraphs of preamble, ChatGPT is less likely to extract and cite that answer. State the fact directly: "SEO services in Canada typically cost between $1,500 and $5,000 per month, depending on scope and competition."
Stay internally consistent. If your prices, your services, your address, or your positioning differ between your homepage, a services page, and a directory listing, the AI has to pick. It often picks wrong, or picks a competitor that doesn't contradict itself.
Consistency operates across multiple dimensions:
- Business name (exact spelling and format)
- Address (same format everywhere)
- Service descriptions (identical across platforms)
- Pricing (current and consistent)
- Business hours (accurate and updated)
Our guide on AI trust signals explains why consistency is the most underrated factor in AI visibility.
Make it machine-readable. Schema.org structured data — Organization, Service, FAQPage, BreadcrumbList — gives extraction a map. Without it, the AI is reading your HTML like a visitor squinting at a messy page.
Our guide on structured data for AI visibility covers the specific schema types that matter most for ChatGPT citations.
The citation loop nobody notices
The single most underrated lever is consistency multiplied by repetition across the web. When the same facts about you appear on your site, your reviews, your case studies, and reputable third-party sources, an AI engine triangulates: the repeated, stable version is the trustworthy one.
This sounds like old-school brand consistency, and it is — but the payoff is different now. In the link era, consistency built ranking. In the answer era, consistency decides whether the AI writes your name into the citation, or chooses someone else's.
The citation loop works like this:
- Consistent information across your website and third-party sources
- AI engines identify your entity with confidence
- Your content gets cited in AI answers
- Citations create more references to your brand
- More references strengthen your entity signals
- Stronger entity signals lead to more citations
Each iteration of this loop makes the next one more effective. This is why early investment in consistency pays compound returns.
The extractability test
Before you publish any page, run this test: paste your page URL into ChatGPT and ask it the question your page answers. Look at the answer it gives. Look at the citations underneath.
If your page is cited, check whether the answer it pulled matches what you intended. If it pulled the right information, your page is extractable — the content structure is working. If it pulled something else or didn't cite you at all, your page needs work.
The most common failure modes in the extractability test:
- No citation at all — The AI answered the question without referencing your page. Usually means your content isn't structured for extraction, or your entity signals are too weak for the AI to connect your page to your brand.
- Wrong information cited — The AI cited your page but pulled incorrect or outdated information. This means your page has consistency issues — the right answer exists but is buried below contradictory content.
- Partial citation — The AI cited your page but only used a fragment, missing the key insight. This usually means your opening paragraph doesn't lead with the direct answer.
Run this test for your five most important pages. The results will tell you exactly where to focus your optimization efforts. Our guide on AI visibility audit provides the complete framework for systematic testing.
Common mistakes that keep sites out of ChatGPT answers
After auditing hundreds of brands, we see the same mistakes repeatedly:
-
Burying answers — The information exists but is hidden below preamble, navigation, or design elements. AI engines need clear, extractable content.
-
Inconsistent business information — Your website says one thing, your Google Business Profile says another, and your directory listings say something different. This inconsistency makes AI engines uncertain about your identity.
-
Missing structured data — Without schema markup, AI engines have to guess at your entity information rather than reading your explicit declarations.
-
Generic content — Pages that repeat common advice without adding unique insight. AI engines prefer sources with original data, named methods, or first-hand experience.
-
Ignoring cross-web presence — Focusing only on your website while ignoring directory listings, reviews, and social media profiles that AI engines use to corroborate your identity.
-
Not updating content — Publishing a page and never refreshing it. AI engines prefer current sources, and stale content signals neglect.
Our guide on entity SEO for AI search covers building comprehensive entity signals.
A content template that earns ChatGPT citations
Here's the specific format we've found works best for earning ChatGPT citations:
Opening (2-3 sentences): State the direct answer to the question. Include the key fact, number, or recommendation in the first sentence. This is what ChatGPT extracts.
Context (1-2 paragraphs): Expand on the answer with supporting details, factors that affect the answer, and specific examples. This gives the AI depth to work with if it wants to elaborate.
Specifics (bulleted list): Break down the key points into clear, extractable bullets. Each bullet should be a standalone fact that could be quoted independently.
Proof (1 paragraph): Reference specific results, data, or case studies that validate your answer. This is the authority signal that tips the scales.
This format works because it mirrors how ChatGPT constructs answers: it pulls the direct statement, adds supporting detail if needed, and cites the source that provided the clearest, most complete version of the answer.
What to actually do this quarter
-
Run your three highest-intent buyer questions through ChatGPT search right now and screenshot the citations. That's your baseline.
-
For each question where you're missing, write one page that answers it in the first two sentences, in plain language, with explicit numbers and facts.
-
Fix contradictions. Search your site for the different ways you describe your service area, your process, and your proof points, and make them say the same thing.
-
Add the schema that describes your business as an entity, not just as pages. Our guide on structured data for AI visibility covers implementation.
-
Re-run the same questions in sixty days and measure the citation delta.
The brands that get ahead here are the ones that started treating "ranked in ChatGPT" as "named as the source". The window is still open because most competitors are still optimizing for ten blue links that fewer and fewer people ever scroll to.
For a deeper dive, see our guide on understanding your AI visibility score.
For a deeper dive, see our guide on how to audit your AI visibility.
For the full playbook on building AI recommendation signals, see our guide on how to get recommended by AI.
Want your own baseline? We run a citation audit across ChatGPT, Perplexity, Gemini, and AI Overviews as part of our teardown — book a free website teardown and we'll show you exactly where the AI answers are ignoring your business.