Guides
When a buyer asks an AI assistant for a supplier, a few businesses make it into the answer and most do not. Here is what happens between the question and the answer, and the signals you can actually work on.
Search First, Then Write
With web search turned on, as it is by default for many questions, ChatGPT, Perplexity and Gemini follow roughly the same steps:
- They turn the buyer’s question into one or more web searches.
- They read a small number of the pages they find.
- They write an answer from what they read, and often list the pages they relied on as sources.
Two things follow. First, to be named you usually need to be found and read in that short search step. Second, the assistant names businesses it can justify from what it read: a clear statement of what you do, for whom, and proof that you do it well.
Without web search, an assistant answers from what it learned during training. Businesses that are widely and consistently described across the web have an advantage there too.
The Signals That Decide Who Gets Named
1. The assistant can read your site
If your robots.txt blocks AI crawlers, if your content only appears after JavaScript runs, or if important pages are missing from your sitemap, the assistant may simply never see them.
2. It understands who you are
Assistants need to know what your business is, what it offers, to whom and where. A vague homepage and no “About” page make that hard. A clear description, consistent company name and structured data (schema.org Organization) make it easy.
3. You have a page that answers the question
If a buyer asks about lead times for small batches and no page on your site talks about lead times or small batches, the assistant will quote someone who does. Often the gap is simply a missing page.
4. The answer is direct
Assistants quote sentences that answer the question plainly. An answer buried in the fifth paragraph of a sales page is harder to use than a short, direct statement near the top, under a heading that matches the question.
5. You show proof
Prices or price ranges, lead times, certifications, capacities, case studies and figures give the assistant something concrete to repeat. “Fast delivery” is a claim; “parts shipped in 5 working days for batches under 500” is evidence.
6. Others talk about you
Assistants also cite directories, comparison articles, reviews and trade publications. Being present and described accurately in those third-party sources often decides whether you are named.
7. Your content is current
When two sources say similar things, recent and dated information has an edge. Old prices, past years in titles and outdated certifications work against you.
Why the Engines Disagree
ChatGPT, Perplexity and Gemini rely on different search sources and different models, so they rarely name exactly the same businesses. The same engine can also give a different answer to the same question an hour later. That is why one-off checks mislead, and why measurement needs several engines, repeated answers and a view over time.
Read the Landscape Before You Act
Looking at who wins each question tells you what is realistic:
- Nobody is cited: the assistant improvises. The first business with a clear, direct answer has a real chance to become the reference.
- An official source dominates (a standard, a public body, a large reference site): you are unlikely to replace it. Aim to be the secondary source, or target the more specific variants of the question.
- Competitors are cited: study the pages they win with, and build a better answer with stronger proof.
Ready to act? Our 12-point checklist turns these signals into concrete fixes.
