引耀
中文 Free sample

How to check if ChatGPT recommends your SaaS

Updated

You can check this without us. The point is to make the check something you can repeat.

Steps

  1. Write 10 to 20 questions a buyer would actually ask. Include “best X for Y” questions that do not contain your name, and one or two that do. Questions that use the name answer a different question: “what do you think this company is?”
  2. Pick one mode and stick to it. Our published run was ChatGPT on chatgpt.com, logged out, with the experimental chatgpt-web engine in citeshine-audit (MIT). Logged-out web answers can differ from API answers and from a logged-in account.
  3. Run each question more than once if you want a rate. Our day-0 report is 13 prompts × n=1: 13 successful, 0 errors. We do not treat that as a stable rate. The report says so.
  4. For each answer, record whether you were mentioned, the order when you were listed, the source labels, and any sentence that describes you incorrectly. Keep the date and the engine next to the number.
  5. Read the wrong sentences before you celebrate a mention. On our run, both mentions described a Shopify store the domain used to be. Being named was the problem, not the win.

The public file is the day-0 report. Raw answers were not published. The most-cited-domains table is (none) because that export had no citation URLs.

If you want us to run it

The form asks for a website and an email. Competitors and questions are optional. We confirm the question set before we start. The free sample is close to this page: ChatGPT only, up to 5 buyer questions, and a short summary. Starter is $99 for ChatGPT, Perplexity, Gemini, DeepSeek, Doubao and Kimi, up to 10 questions, a competitor comparison, cited sources, and a written fix list. We have not published a monthly price for the work after the audit.

FAQ

Which engine did you publish?

ChatGPT web, logged out, on 2026-10-13, 13 prompts, n=1. API engines for OpenAI, Perplexity, DeepSeek, Kimi, Grok, Claude and Gemini exist in citeshine-audit and have not been used in a published run. Doubao web support is on the tool's roadmap.

Is one answer enough?

No. Our published baseline is n=1 and we label it that way because the variance is high. For a check you rely on, run each question more than once and report the rate.

What if I would rather not run it?

Send the website and your email. The free sample is ChatGPT only, up to 5 buyer questions, and a short summary. Starter is $99 for all six engines, up to 10 questions, a competitor comparison, cited sources, and a written fix list.

Start with a free sample check

ChatGPT only, up to 5 buyer questions, and a short summary. Upgrade to the full $99 audit when you want all six engines and the written fix list.

Request the free sample Compare the two Request the $99 Starter