Key Takeaways
- In 15 minutes and with free tools, you can see whether AI assistants cite your site and whether anything is blocking their crawlers.
- Search crawlers (OAI-SearchBot, PerplexityBot, Bingbot, Googlebot) decide whether you can be cited; training crawlers such as GPTBot do not.
- Since 15 September 2026, Cloudflare's new defaults can block multi-purpose crawlers like Googlebot and Bingbot if Training is blocked. Check your setting.
- Bing Webmaster Tools' AI Performance report shows real Copilot citations by page; GA4 shows the visits AI answers send you.
Picture a marketing team with 40 service pages that rank well on Google, yet the company never appears when a prospect asks Copilot for a supplier shortlist. Often the cause takes minutes to find: a firewall or CDN rule is turning away Bingbot.
What does an AI citation self-check tell you?
It tells you three things: whether AI assistants already mention or link your site for the questions your buyers ask, whether your robots.txt, firewall or CDN is blocking the crawlers those assistants rely on, and whether AI answers are sending you any measurable traffic. Each answer points to a different fix.
Being cited is different from ranking. An assistant can answer a question about your category without naming you, even if you rank on page one, because it only cites pages its own search crawler fetched and could parse.
What do you need before you start the AI citation check?
You need a browser, access to your website's robots.txt and CDN or firewall settings, a Google Analytics 4 property, and a free Bing Webmaster Tools account with your site verified. Gemini, Copilot and Perplexity all work in Hong Kong on free plans; ChatGPT is not officially offered here.
| Tool | What it shows | Where to look | Usable in Hong Kong? |
|---|---|---|---|
| Gemini | Whether answers mention or link you | gemini.google.com, ask with sources shown | Yes |
| Copilot | Whether answers cite your pages | copilot.microsoft.com | Yes |
| Perplexity | Numbered source list per answer | perplexity.ai | Yes |
| Bing Webmaster Tools | Copilot and Bing AI citations by page | AI Performance report (public preview) | Yes |
| Google Analytics 4 | Visits from AI assistants | Reports > Acquisition > Traffic acquisition | Yes |
| ChatGPT | Whether ChatGPT search cites you | chatgpt.com | Not officially supported |
How do you test whether AI assistants cite your website?
Ask each assistant five questions a real buyer would type, in a fresh chat with no history, and record whether your brand is mentioned, whether your URL is linked as a source, and which competitors appear instead. Use the same five prompts every month so the results are comparable.
Step 1: Write five buyer questions (3 minutes)
Take them from sales calls or enquiry emails, not from your own keyword list. Mix one category question, one comparison, one price question, one local question and one problem question.
Prompt template (paste into Gemini, Copilot and Perplexity, one at a time):
I run [type of company] in Hong Kong. Which [service or product] providers should I shortlist for [specific need]? Give me 5 options with a one-line reason each, and list the sources you used with links.Follow-up prompt (same chat):
What do you know about [your brand name] and its [service]? Which pages did you read to answer? If you could not find information, say so.Step 2: Log the result (2 minutes)
For each answer write down: mentioned (yes/no), linked (yes/no and which URL), and the competitors named. If the follow-up prompt returns outdated or wrong facts, note the page the assistant cites; that page is what you fix first.
How do you check that robots.txt and Cloudflare are not blocking AI crawlers?
Open yourdomain.com/robots.txt and confirm that OAI-SearchBot, PerplexityBot, Bingbot and Googlebot are not disallowed. Then check your CDN: on Cloudflare, review the AI crawler settings under Security, because blocking Training crawlers there can also block Googlebot and Bingbot.
Step 3: Read robots.txt (3 minutes)
Each assistant uses its own search crawler, and that is the one that decides whether you can be cited. According to OpenAI's crawler documentation, sites that block OAI-SearchBot will not be shown in ChatGPT search answers, while GPTBot only controls training. Perplexity's documentation says the same about PerplexityBot, which is not used for model training.
# A robots.txt that allows AI search but opts out of AI training
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: GPTBot
Disallow: /
User-agent: *
Allow: /
Sitemap: https://[yourdomain.com]/sitemap.xmlGoogle works differently. Google-Extended is a robots.txt token, not a crawler, and it controls whether content is used to train and ground Gemini models. Google states it is not a way to manage how content appears in Google Search, so blocking it does not remove you from AI Overviews. Blocking Googlebot does.
Step 4: Check your CDN or firewall (3 minutes)
On Cloudflare, open your domain and review the AI crawler controls in the Security settings. Cloudflare announced that from 15 September 2026, new domains block Training and Agent crawlers by default on pages that show ads, and that multi-purpose crawlers such as Googlebot, Applebot and Bingbot are blocked when Training is blocked, even if Search is allowed.
How do you measure AI citations and AI referral traffic each month?
Use Bing Webmaster Tools' AI Performance report for Copilot and Bing AI citations by page, and a GA4 traffic filter for visits from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. Citations show visibility; referral sessions show whether those answers actually send people to you.
Step 5: Open the Bing AI Performance report (2 minutes)
In Bing Webmaster Tools, verify your site, then open AI Performance in the left menu. The report, in public preview since February 2026, lists how often your pages are cited in Microsoft Copilot and Bing AI summaries, by page and by grounding query. It does not cover ChatGPT, Perplexity or Google, and it has no click data yet. Allow about 24 hours after verification for data to appear.
Step 6: Filter AI referrals in GA4 (2 minutes)
In GA4 go to Reports > Acquisition > Traffic acquisition, set the primary dimension to Session source, then add a filter with the regex below. Save it as a comparison so it is one click next month.
GA4 filter, Session source, matches regex:
.*(chatgpt\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|bing\.com/chat).*Note that clicks from Google AI Overviews arrive as normal Google organic traffic.
How much does an AI citation self-check cost?
The check itself costs nothing. Every tool in this guide has a free tier, and each run takes about 15 minutes. The real cost is time: one person, once a month, roughly three hours a year, plus any developer time needed to change robots.txt or firewall rules.
Paid AI visibility trackers automate the prompt testing across hundreds of questions. They make sense once you know your site is readable; buying one while Googlebot is blocked only gives you a dashboard of zeros.
What usually goes wrong, and how do you fix it?
Three problems cause most missing citations: crawlers blocked by a CDN or firewall default, key content that only appears after JavaScript runs, and citations concentrated on one old page. Each has a specific fix you can test the same day by re-running the prompts after the crawler revisits.
Problem 1: A security default blocks search crawlers
Symptom: you rank on Google but never appear in Copilot. Fix: allow Search crawlers explicitly, and if you block Training on Cloudflare, check whether Googlebot and Bingbot are affected, as Cloudflare's own announcement warns. Re-test after a few days.
Problem 2: Your content needs JavaScript to appear
Vercel's study with MERJ found GPTBot made 569 million requests across its network in one month, yet none of the major AI crawlers, including OAI-SearchBot and ChatGPT-User, rendered JavaScript at the time (December 2024). Some site owners report changes in 2026, but do not rely on it. Fix: right-click, View Page Source, and search for your key prices and service names. If they are missing from the raw HTML, ask your developer for server-side rendering on those pages.
Problem 3: One outdated page gets all the citations
In a Search Influence analysis of about 20,000 Copilot citations, one page took 69% of the total. If your most-cited page has old prices, every AI answer repeats them. Fix: sort the Bing AI Performance pages by citations, update the top page first, and add a clear one-paragraph answer near the top of each key page.
Your 15-minute AI citation checklist
- Five buyer prompts tested in Gemini, Copilot and Perplexity, fresh chat, results logged
- robots.txt allows OAI-SearchBot, PerplexityBot, Bingbot and Googlebot
- Cloudflare or firewall AI crawler settings reviewed after the 15 September 2026 default change
- Bing Webmaster Tools verified and AI Performance report opened
- GA4 AI referral filter saved as a comparison
- Key prices and service names visible in View Page Source
- Most-cited page checked for outdated facts
Run it on the first working day of each month and keep the same five prompts. The trend matters more than any single answer.
Last verified: 10 November 2026 (OpenAI, Perplexity and Cloudflare documentation; Bing Webmaster Tools report scope).
Empowering AI. Centered on you.
Want the full picture in 30 seconds?
The manual check covers five prompts. A free AEO audit scans your site for crawler access, structure and AI readiness in one pass, and shows what to fix first. If you want help acting on the results, we'll walk you through every step.
Frequently Asked Questions
Does blocking GPTBot stop ChatGPT from citing my site?
No. According to OpenAI's documentation, GPTBot only controls training. ChatGPT search citations depend on OAI-SearchBot, which you can allow separately.
Can I use ChatGPT for this test in Hong Kong?
Hong Kong is not on OpenAI's supported list, so use Gemini, Copilot and Perplexity, which all work here on free plans. Your overseas customers may still see ChatGPT answers, so keep OAI-SearchBot allowed.
How long after a fix will AI answers change?
Crawlers need to revisit the page first. Perplexity says robots.txt changes can take up to 24 hours to apply; re-test your five prompts after about a week.