How to get cited by ChatGPT: 7 levers

How ChatGPT picks sources

1. 01Knowledge or searchFor general questions ChatGPT uses its training knowledge. For current or local questions it searches the web.
2. 02Collect candidatesPages from the search results make the shortlist. Pages that rank high are more likely to be included.
3. 03Build the answerChatGPT takes suitable passages, writes the answer and names some of the sources.

Being in the index is not enough. ChatGPT has to match your content to the right question and rate it as a suitable source. This analysis shows how much the Google ranking counts:

In ChatGPT the share drops from about 30% in position 1 to about 12% in position 3. Websites further down are rarely named. If you want to appear in ChatGPT, you need good Google positions first.

For which questions ChatGPT names providers

Not every question leads to a recommendation. For knowledge questions ChatGPT mostly explains. When people look for providers, it almost always names businesses:

For you this means: first check which questions your customers ask. Work on ChatGPT pays off most for provider and price questions.

The 7 levers

Each lever says whether it was measured in our data or is based on experience. Note: for page features such as direct answers, our data did not yet support a firm statement for ChatGPT. The evidence for those comes from Google AI Overview and Perplexity.

1. 1

   Rank well on Google

   

   ChatGPT often searches through a search engine. What is missing there is rarely named.

   How to do it: Check indexing, loading speed and page structure and improve the pages just short of page one.

   Measured in our dataPositions 1 to 3 are cited in ChatGPT about 3.7 times as often as positions 4 to 10, provider analysis.

2. 2

   Answer the question in the first sentence

   

   ChatGPT takes individual passages. A passage that answers right away is easy to use.

   How to do it: Answer blocks of two to four sentences that answer a question directly and fully, without a preamble.

   Measured in our dataReliable in Google AI Overview and Perplexity. For ChatGPT our data was not enough, data test.

3. 3

   Say clearly who you are

   

   ChatGPT ties statements to a business, a service and a place. Five different names across ten pages make that hard.

   How to do it: Use the same name, services and location everywhere, including your structured data (Schema.org).

   Experience, not measuredFrom our work on entities, without our own measurement of the effect.

4. 4

   Get mentioned on other sites

   

   ChatGPT builds answers from several sources. If you are missing from the pages cited for a topic, you are named less often.

   How to do it: Check which directories, trade media and comparison sites ChatGPT names for your questions, and get present there.

   Experience, not measuredWe see which pages are cited in every measurement. We have not measured the effect of a new mention.

5. 5

   FAQ with real customer questions

   

   An FAQ section helps when every answer comes in the first sentence. FAQ schema alone is not a shortcut.

   How to do it: Collect questions from calls, emails and Search Console and answer each in the first sentence.

   Measured in our dataAhrefs found no rise in ChatGPT citations across 1,885 pages that added FAQ schema (Ahrefs 2026).

6. 6

   Keep it current

   

   For current questions ChatGPT searches the web. Outdated prices or services make poor sources.

   How to do it: Show an update date on important pages and visibly keep the content up to date.

   Experience, not measuredIn the data test, “updated in the last year” went with citations in Google’s AI Overview, but not reliably. There is no statement for ChatGPT.

7. 7

   Measure repeatedly

   

   ChatGPT does not answer the same way every time. A single query says little.

   How to do it: Ask the same questions on several days and count your mentions, separately for each service.

   Measured in our dataAsked the same question three times, ChatGPT made the same call on naming providers in about 93% of cases. Which providers it names varies more.

What does not help, or barely

• Relying on llms.txt: the file is widely discussed. We know of no evidence that it makes ChatGPT cite you more.
• Piling up keywords: overloaded texts rarely have a clear answer in the first sentence.
• Format rules as a trick: lists, tables, statistics and short sentences showed no reliable link with citations in any service in our data.
• Quick tips from social media: they change all the time. What lasts is ranking, clear answers, mentions and measurement.

Checklist in 10 points

• Do you know the questions for which ChatGPT names providers in your industry?
• Are your most important pages on page one of Google?
• Does every page say clearly who you are, what you offer and where?
• Does every important page answer its core question in the first paragraph?
• Are there answer blocks of two to four sentences?
• Are there FAQ sections with real customer questions, answered in the first sentence?
• Are you present on the pages ChatGPT cites for your questions?
• Do important pages show a visible update date?
• Do you measure mentions repeatedly and separately for each service?
• Do you know your baseline before you change anything?

Do it yourself or get help?

Clear answers, FAQ and consistent details are things you can do yourself. For ongoing measurement across several services and the work on mentions on other sites, tools and experience help. If you hand it to us, we give you a fixed price based on scope before we start.