How to check whether AI assistants cite your site (a developer's workflow)

Search traffic used to mean one thing: Google results. Now a growing share of questions go to AI assistants like ChatGPT, Perplexity and Gemini. If you ship a product or docs site, it's worth knowing whether those assistants mention you.

This post walks through a simple, repeatable way to check. It needs a spreadsheet, a bit of discipline and, optionally, a small script.

Don't test with your brand name. Test with the problems your users have.

Aim for 10 to 20 prompts. Mix broad ones ("best tool to monitor cron jobs") with specific ones ("how do I alert on a missed cron run in Node").

Use one row per prompt per assistant per check. These columns are enough:

date | assistant | prompt | brand _ mentioned ( y / n ) | our _ url _ cited ( y / n ) | cited _ urls | competitors _ named | notes Enter fullscreen mode Exit fullscreen mode Record the cited URLs, not only whether you appeared. The list of cited domains is the most useful part. It shows which pages assistants trust for your topic.

AI answers vary between runs, accounts and locations. You can't remove that noise, but you can keep it steady.

Assistants that browse the web, like Perplexity, show sources directly. For others, ask a follow-up such as "which sources did you use?" and note what you get.

If an assistant offers an API, you can loop over your prompts and save the raw answers. API answers can differ from the consumer chat apps, so treat them as a signal, not ground truth.

import csv, datetime, re def extract_urls ( text ): return re. findall ( r " https?://[^\s)\]]+ ", text ) def log_answer ( writer, assistant, prompt, answer, brand, domain ): urls = extract_urls ( answer ) writer. writerow ([ datetime. date. today (). isoformat (), assistant, prompt, " y " if brand. lower () in answer. lower () else " n ", " y " if any ( domain in u for u in urls ) else " n ", " ". join ( urls ), ]) # answers = {prompt: call_your_assistant_api(prompt) for prompt in prompts} # with open("ai-checks.csv", "a", newline="") as f: # w = csv.writer(f) # for p, a in answers.items(): # log_answer(w, "assistant-name", p, a, "YourBrand", "yourdomain.com") Enter fullscreen mode Exit fullscreen mode 5. Read the results like a developer Look for patterns, not single answers.

A small, steady checking habit beats guessing. Twenty prompts, a sheet and a fortnightly rerun will show you more than any single screenshot.

I work on Serpair, which runs this kind of Google and AI-answer tracking for you. But the manual method above works fine on its own, and it's a good place to start.