Getting found in ChatGPT

How does ChatGPT choose what it names?

ChatGPT builds its answer from two things: what the model has learned, and sources it looks up live. For brand mentions it looks at sources that are relevant, clear and reliable for the question. Authority, recency and structure all play a part. Brands that consistently come up in those sources are named more often.

Important: the answer differs per question, per language and per moment. That is why checking once is not enough, and measuring over time is the only way to know where you stand.

Step 1: make sure you may be crawled

If your robots.txt blocks the AI crawlers, you can never end up in an answer. This is the most common mistake and the easiest to fix. Check that these crawlers have access:

Step 2: write quotably

Assistants like to take over short, factual pieces. Make it easy for them:

• Put a core answer of two to three sentences at the top of every page.
• Use clear question headings ("What is...", "How does... work").
• Work with lists, steps and tables instead of long walls of text.
• Add an FAQ with the literal follow-up questions.
• Back it up with figures and a date, so the source looks current and concrete.

Step 3: build authority on the right sources

ChatGPT draws on sources it trusts. Make sure you are there: trade media, category overviews ("best X"), reliable references and your own strong pages. A mention in an authoritative overview often does more than ten separate pages. Consistent information about your brand (name, what you do, for whom) helps the model place you correctly. Which sources those are for your market is shown by a per-question measurement; more about that on improve.

Step 4: measure whether it works

Without measuring you do not know. Put the questions your customers ask to ChatGPT repeatedly and count how often you are named, alongside which competitors, and from which sources. Do that over time and with an interval, so you can tell a real rise from chance. Geony automates this, across multiple assistants at once.

Does this also work for Gemini, Claude and Perplexity?

Yes, the basics are the same: being allowed to be crawled, writing quotably, authority on the right sources. But the outcome differs strongly per assistant. A source that counts with ChatGPT often does not count with Gemini, and Perplexity shows its sources more explicitly than the rest. That is why you best measure them all separately, instead of relying on one. See also GEO optimization for the broader approach.