ChatGPT cites a consultant in two situations: when it searches the web while answering and finds a page on that consultant's site that answers the question more specifically than anything else it can reach, or when someone asks who to hire and the firm appears in what it reads. You improve your chances by letting AI crawlers into your site, publishing many narrow, well-sourced articles on the exact questions your clients ask, and measuring retrieval in your server logs instead of waiting for clicks that rarely arrive.
Nobody can guarantee a citation. What you can control is whether your pages are reachable, worth quoting, and numerous enough to be read often.
When does ChatGPT actually cite a consultant?
Less often than most consultants assume, and in a narrower set of situations.
On 23 September 2026 we put 20 German business questions to the model, the kind a tax adviser, HR consultant or compliance specialist answers for clients every week. In 10 of them the model answered without searching at all and cited nobody. Where it did search, it cited primary sources (laws, ministries, the OECD, large accounting firms) on 9 of the 10. Consulting firms were named only on questions about who to hire or how to choose one. And when we ran the same question twice, it named two different sets of firms.
| Situation | What the model did in our test | What it means for a consultant |
|---|---|---|
| The model did not search | Answered from its own knowledge, cited nobody | No website gets credit; you cannot win these answers directly |
| The model searched | Cited primary sources on 9 of 10 questions | Your page competes with the statute and the ministry, so it must add application, examples and edge cases they do not |
| Who should I hire, how do I choose | The only questions where firms were named; two runs named different firms | Being present in the material it reads matters, and one appearance is a sample, not a position |
So the realistic goal is not to be quoted on every question in your field. It is to be the most useful page the model can find on the specific, applied questions where it does search, and to be present when a buyer asks for recommendations.
What does your site need before ChatGPT can cite it?
Access comes first. ChatGPT reaches the web through OpenAI's agents: GPTBot crawls broadly, while OAI-SearchBot and ChatGPT-User fetch pages when ChatGPT is searching or answering a person. If your robots.txt, firewall or CDN turns them away, nothing you write will be read.
Consultant sites are especially prone to this because many run on hosted website builders or behind a CDN with bot protection switched on by default, a setting the owner never chose. Check your robots.txt against the steps in how to allow GPTBot in robots.txt, and if you use Cloudflare, read GPTBot blocked by Cloudflare. A quick manual test: open yourdomain.com/robots.txt and look for Disallow lines under GPTBot, OAI-SearchBot or ChatGPT-User, or a blanket Disallow under "User-agent: *". Then confirm in your server log that those agents are receiving normal responses rather than errors or challenge pages.
After access, the basics:
- Pages that render as readable HTML without waiting for JavaScript.
- A sitemap that lists every article.
- A clear author name, firm name and location on each article.
- Dates on articles, updated when the rules change.
These are necessary. As the next sections show, they are not sufficient.
What kind of content gets a consultant quoted?
Because the model reaches for primary sources when it searches, a consultant's page earns a place by doing what the primary source does not: explaining how the rule applies to a particular kind of client in a particular situation.
Compare two titles. "What is transfer pricing" is answered by the law, the OECD and every large firm, and the model may not search for it at all. "How a small German GmbH with a US parent documents intercompany service fees" is a question a real buyer asks, the primary sources answer it only indirectly, and a specialist's worked explanation is genuinely the best page on it.
What quotable consultant content looks like in practice:
- One question per page. The title is the question, phrased the way a client would say it.
- The answer in the first two sentences. Assistants lift short, self-contained passages. If the answer is buried in paragraph six, it is less likely to be used.
- Link the primary source yourself. The model is already looking for the law or the ministry guidance. Be the page that cites it and then explains what it means in practice.
- Concrete examples and edge cases. The exception that trips people up is where your experience shows and where generic sources are thin.
- Pages that help someone choose. Since firms were named only on hiring questions, publish honest guides on how to choose a consultant in your specialty, including the situations where a client does not need one. This is the material the model reads when someone asks who to hire.
For more on the recommendation side, see how to get ChatGPT to recommend your business.
Why does the number of pages matter more than perfect markup?
We watched a site with complete technical setup (server rendering, Article, FAQPage and Organization JSON-LD, a sitemap) but only 75 URLs stay absent from answers on the exact topic of one of its own articles. Meanwhile a site with about 190,000 crawlable URLs was crawled constantly. Surface area mattered more than markup.
The crawl volumes show how much attention AI crawlers spend on large sites. Across the retained nginx logs of an edtech platform we work with, from 28 July to 11 August 2026, ClaudeBot made 19,272,125 requests, Meta-ExternalAgent 3,854,076, GPTBot 2,744,985, Applebot 2,655,555 and Amazonbot 1,170,268. Googlebot made 52,292. Crawlers go where there is a lot to read and come back often.
A typical consultant site has a homepage, a services page, an about page and a handful of posts from a few years ago. That gives crawlers little reason to return and gives the model little to retrieve. You do not need hundreds of thousands of pages, but you do need a steady body of specific articles that grows month after month, each answering a question your clients actually ask.
How do you know if ChatGPT is reading and using your pages?
Not from your analytics dashboard. In about two weeks of logs from three sites, with roughly 1.2 million requests in the two busiest logs, human visitors arriving from AI assistants totalled one from Copilot and zero from ChatGPT, Perplexity, Gemini, Claude, Meta AI and the rest. Assistants read your page and answer in place. A citation can happen without anyone clicking through, so referral traffic will look flat even when things are working.
The evidence is in your server log. Look for the live retrieval agents: ChatGPT-User, OAI-SearchBot and Perplexity-User. These fetch a page while an assistant is answering a real person. In the same fortnight, the edtech platform received 1,918 such fetches on content pages, and a client blog of a few dozen articles received 56 across 24 different articles. A modest blog can be retrieved during real conversations; the log is how you see it.
Use those fetches as your feedback loop. Note which articles are retrieved, then write more articles on neighbouring questions. If you have not worked with raw logs before, how to find AI crawlers in server logs walks through it.
Asking ChatGPT about your specialty yourself is still useful, but treat it as a noisy sample. Two runs of the same question named two different sets of firms in our test, so run a question several times and look for patterns rather than reacting to a single answer.
What mistakes stop consultants from being cited?
- Blocking crawlers without knowing it. A default CDN or plugin setting can undo everything else.
- Writing brochures instead of answers. Service pages describe what you do; the model is looking for pages that answer what the client asked.
- Restating the rule without applying it. If your page says what the law says, the model will cite the law.
- Publishing a burst and stopping. A few posts do not create the surface area that keeps crawlers returning.
- Judging by clicks. Near-zero referral traffic from assistants is normal and tells you little.
- Paying for a promise. Anyone guaranteeing ChatGPT citations is promising something they do not control.
If you would rather not run the writing and measurement yourself, ItGrows for Business writes and publishes research-grade articles to your own domain, structured so answer engines can quote them, and reports from your server log which AI crawlers read them and which assistants retrieved them while answering someone. It costs $499 a month (plus a one-time $499 setup, free with 3 months or a year), and it does not guarantee citations, because nobody can.
Frequently Asked Questions
Can a consultant pay to be cited by ChatGPT?
No. There is no paid placement inside ChatGPT's answers, and nobody can guarantee a citation. What you can pay for is the work that makes citation more likely: crawler access, a steady stream of specific articles, and log-based measurement of whether assistants retrieve them.
Does schema markup help consultants get cited?
It helps machines parse your pages, and it is worth having. But we saw a site with complete Article, FAQPage and Organization markup and only 75 URLs stay absent from answers on its own topic, while a very large site was crawled constantly. The volume and specificity of your content matter more than the markup.
Why does ChatGPT name different consultants each time I ask?
Answers vary from run to run; in our test two runs of the same hiring question named two different sets of firms. Treat any single answer as a sample, run questions several times, and focus on being consistently present in the material the model reads rather than on one result.
How can I tell whether ChatGPT has used my articles?
Check your server log for ChatGPT-User, OAI-SearchBot and Perplexity-User requests, which are fetches made while an assistant answers a real person. Referral traffic is a poor signal because assistants rarely send clicks. If none of these agents appear at all, check your robots.txt and your CDN's bot settings first, because they may be turned away before they can read anything.