Free sample

Try CitePack free on one question

See what AI engines tell your buyers when they ask one buying question.

  • One buying question. Pick one of three for your category, or write your own.
  • Two of the four AI engines, your pick, each asked three times.
  • Who is named first in each answer, and whether your company appears and where.
  • The other companies named, and the sources the engines cited.

One free sample per company. We email you the link. It covers one question on two of the four engines. A full Decision Map covers five questions on all four engines, with a Work Plan.

See the full Decision Map
How AI chooses vendors

How AI chooses between vendors when a buyer asks.

A buyer asks ChatGPT, Claude, Gemini or Perplexity what to buy and gets a few names back. Before it names anyone, the engine has already done some work. This page covers what that work is, how much of it you can see, and how to read the trail for your own category.
On this page

What does an AI engine do before it names a vendor?

With web search on, it does three things: it searches, it looks up a shortlist of companies by name, and it reads pages. The names in the answer come after that trail.

  • It searchedThe searches it ran, which can be kept word for word.
  • It looked up a shortlistThe companies it looked up by name.
  • It read pagesThe pages behind the answer, such as lists, company pages and forum threads.

Not every answer shows every move. Where an engine doesn’t show a step, that step is unknown, not zero.

Why doesn’t one AI answer tell you where you stand?

Because the same question can get a different answer the next time you ask it. In the app, each buyer’s screen is also shaped by their own history and settings. One screenshot describes one person. To see a pattern, the same question has to go to each engine more than once, with clean repeats and the same settings every time.

Read how each engine is asked

Is being mentioned the same as being recommended?

No. An answer can mention a company without offering it, and it can cite a page without recommending that page’s product. Being put forward as an option, being mentioned and being cited are three different facts.

Order is a fourth. When an answer lists options in a clear order, you can read who came first. When it doesn’t, no one came first, and guessing would be wrong.

Can you tell why an engine chose one vendor over another?

Not from the outside. You can see what an engine searched, which companies it looked up and which pages it read. Its reasoning stays out of view, and nothing should claim otherwise.

What you can do is set the trail beside the answer: whose pages it read, who it put first, and where your own pages were used or left out.

How do you see this for your own category?

CitePack is built for this. It maps one defined decision at a time: your offering, the buying category it’s bought in, and your buyers’ market. The category and the buyer’s situation shape which names an engine puts forward. So CitePack frames the questions from the real situations your buyers face, five of them across the whole decision. You give the final go. Each goes to ChatGPT, Claude, Gemini and Perplexity three times, every repeat clean, with your market set on every engine by default. That makes 60 observations of one decision.

  1. 5buying situations, framed by CitePack
  2. × 4engines: ChatGPT, Claude, Gemini, Perplexity
  3. × 3clean repeats on each engine
  4. = 60observations of one decision, every answer kept

The Decision Map shows who each engine put forward, what it said and which pages it read. Beside it, the Work Plan ranks the work those answers point to. Map each offering, category and market you sell into, and each map adds to the picture of where you stand.

See it on an illustrative map: the Crosstrow example map

Read the methodology →See what the Work Plan holds →See pricing →