Before you read another word, try this. Open ChatGPT and type "which [your category] would you recommend in [your country]." Look at the answer. Is your company in it? That little test tells you more about your 2026 marketing than any rank tracker, because 63% of B2B software buyers use ChatGPT to research software (G2, 2026). If you didn't show up, you didn't lose the deal. You were never in the conversation.
The good news is that this is measurable, and it's fixable. Let's walk through how to check it properly and what to do with what you find.
How do you check if AI mentions your brand?
Run the same buying question a real customer would ask, and run it in more than one place. A quick version takes about five minutes:
- Open ChatGPT, Perplexity, and Google's AI results side by side.
- Ask each one to recommend a vendor in your category and market. Use plain language, the way a buyer would.
- Note whether you appear, where you land in the list, and who you're sitting next to.
- Ask a follow-up: "compare the top three and tell me which to pick." See who wins, and why.
Do this a few times with slightly different wording. One prompt is an anecdote. A handful of prompts starts to look like data.
Why isn't your company showing up?
When a brand is missing from AI answers, it's almost always one of three leaks.
Your identity is inconsistent. If your site, your Google profile, LinkedIn, and directories describe you three different ways, the model gets unsure who you are, and an unsure model leaves you out.
Your content isn't answer-first. If the answer to a buyer's question is buried in paragraph eight, the model can't lift it cleanly. It rewards pages that put the answer up top.
Nobody cites you off your own site. More than 80% of the pages AI cites in 2026 don't rank in Google's top 10 (erlin.ai). They live on third-party sources the model trusts, like Reddit, G2, and the press. If you're invisible there, you're invisible in a lot of answers.
Is showing up in ChatGPT the same as ranking on Google?
No, and that trips a lot of people up. The overlap is small. ChatGPT and Claude share only about 8% of the sources they cite (Profound, 2026), and most of what AI quotes doesn't sit in Google's top 10 at all (erlin.ai). Ranking still matters, but a number one spot on Google is no longer a guarantee that you're in the AI's answer. They're related fights with different referees.
How do you get AI to recommend you?
There's no single trick, but there is an order that works. Roughly:
- Add schema markup, and put real data inside it. Generic markup predicted nothing on its own, while Product or Review schema carrying price, aggregateRating and specifications was cited at 61.7% against 41.6% (Fischman, 2026).
- Put quotes and claims next to a named expert, so a specific person can be checked. Citing sources, adding quotations and adding statistics is worth 30 to 40% more visibility in generative answers (Aggarwal et al., 2024).
- Publish your own data. Original numbers are the one statistic the model cannot produce on its own, which is exactly why they give it a reason to name you.
- Keep it fresh, with a real dateModified. Live-web engines read that field when they decide whether your page still answers a time-sensitive question.
- Make your brand identity consistent everywhere it lives.
- Only then worry about llms.txt. Google said in May you don't even need it for AI Overviews.
How long does it take?
It depends on the engine. Live-web tools like Perplexity and Gemini can reflect changes in hours or days. The training-based ones like ChatGPT and Claude take months to catch up when they're working from cold memory. Anyone promising you the top spot in every engine in 30 days either doesn't understand the mechanics or isn't being straight with you.
Start with a number
You can guess at all of this, or you can measure it. We built the AI Visibility Scan for exactly that: a free read of whether AI recommends you across six engines, so your next move is based on data instead of a hunch. Run yours, see where you stand, and go from there.
