How your business shows up on ChatGPT when someone asks about your service
Analysis of 700,000 conversations on ChatGPT shows what kind of content gets cited and what gets ignored. Only 11% of sites are cited by both ChatGPT AND Perplexity. Here's how to get on that list.
Open ChatGPT. Type "recommend me [your service] in [your city]." If your brand doesn't show up, the same question is being asked by your prospects every day, and the answer they get is a list of your competitors. No Google ranking fixes that: it's a different game, with different rules.
The next question is: what makes ChatGPT mention one business and not another? The answer, by the end of 2025, is no longer a mystery. There are data studies on how AIs pick sources, and the patterns are consistent.
What we know about ChatGPT's behavior
A TryProfound analysis covering 700,000 conversations with web citations from Q4 2025 revealed three key findings:
- The first response captures most of the citations. ChatGPT cites sources mostly on its first answer to the prompt, not in follow-up clarifications. So the sites that show up in that first pass are the ones that win visibility.
- Wikipedia, government domains, and authority publications dominate. But the space for private brands is available, especially in niches those sources don't cover (specific B2B categories, local services, new products).
- Cross-platform overlap is low. According to The Digital Bloom, only 11% of sites cited by ChatGPT are also cited by Perplexity. Winning on one platform doesn't guarantee winning on another.
That last part changes the strategy. "Showing up in AIs" isn't enough. You have to measure platform by platform and optimize them separately.
What kind of content ChatGPT cites
From the same studies, the practical pattern:
- Statistics with source. Pages with 5+ concrete, cited data points get mentioned roughly 3 times more than pages without them.
- Self-contained sentences. Sentences that make sense read alone, without prior context. These are what the AI literally "copy-pastes" into its answer.
- Expert quotes. Quotes attributed to people with authority in the topic.
- 120-to-180-word sections. Very short sections (under 50 words) get cited 70% less (SE Ranking). Very long sections without clear structure also lose.
What consistently gets left out: loose narrative prose, long paragraphs without hard data, claims without links, overly promotional content.
How your business shows up, technically
When someone asks ChatGPT about your category, four things happen in sequence:
1. ChatGPT decides whether it needs web search
Only 31% of prompts on ChatGPT trigger a web search; 69% are answered from the model's parameters (Nectiv). If the AI thinks it already knows the answer, it doesn't hit Bing. So you need to be in the model's training data (history) and in the sources Bing/Google index when it does decide to search.
2. If it searches, it uses its retrieval engine
ChatGPT mainly uses Bing as its underlying search engine. That's why having your site correctly indexed in Bing Webmaster Tools is as important as Google Search Console. Many SMBs ignore Bing because its market share is small, but it's the feeder of ChatGPT.
3. It synthesizes the answer
The model receives fragments of several sources and combines them into a single conversational paragraph. Sources that end up well-cited (with a URL visible to the user) are the ones that won the visibility. The rest stay in the internal prompt and aren't seen.
4. The user clicks (or doesn't)
If your brand is mentioned without a link, you only gain brand awareness. If it's mentioned with a link, you gain high-intent traffic. Those visitors convert better than classic organic traffic because they already passed a pre-approval filter.
What actually moves the needle
Translated into concrete actions, the five that move the most:
- Index on Bing. Not just Google. Bing Webmaster Tools has a GSC import, takes 10 minutes.
- Publish content with hard data. At minimum 3 to 5 statistics with source per article, explicit dates, quotes from verifiable sources.
- Structure with clear headings. H2 and H3 framed as questions or direct answers. ChatGPT loves extracting from H2s that look like questions.
- Full schema markup. Organization, Article, FAQPage, LocalBusiness as applicable. You hand the crawler the map of your entity.
- Mention domain entities. If you write about gyms, mention Crunch, Smart Fit, Sports World as reference. The AI learns category-brand associations through co-occurrence.
How we measure this on our projects
At Camino al Sol we run a standardized test every two weeks: 15 prompts like "recommend me private tours in Teotihuacán," "where to book pyramid experiences," with 3 variations each. We run it on ChatGPT, Claude, Perplexity, and Gemini. We graph how many times the studio appeared in the answers, at what position, and whether with a link or just text.
That's the scorecard. At Connect Gyms the test is different (queries about gym SaaS in Spanish and English), but the principle is the same: one hard number every two weeks, comparable over time.
What you can do this week
Open ChatGPT and Perplexity. Type 3 different prompts asking about your category. Write down: do you appear? who does? why them and not you? That 15-minute exercise is the best free GEO diagnosis there is.
If your brand doesn't show up in any answer, write to us. Within 48 hours we'll send you a first audit with three concrete actions and the impact we expect.