Consultant SEO for ChatGPT Search: Get Your Practice Recommended

You cannot buy or guarantee a recommendation, but you can make your practice easy to identify and worth citing: a clear page saying who you help and with what, case studies with real outcomes, reviews on sites AI tools read, and crawling left open. Then test with a fixed set of prompts every month.

The reason this matters to a consultant is that buyers now ask ChatGPT, Google's AI Overviews or AI Mode for "a specialist in supply-chain finance for mid-sized manufacturers", and they get a short list. The list is built from what the tools can find and trust about you. If your site says "I help businesses grow", there is nothing to match. If it says who you serve, what you deliver and what happened for clients, you give the tools something to quote. None of this is a trick; it is mostly clear writing, plus a few checks on technical access.

Follow me on Instagram@sagnikteaches

How the systems pick who to name

I will keep this to what vendors say themselves. OpenAI's crawler documentation describes OAI-SearchBot as the crawler that surfaces websites in ChatGPT's search features: a site opted out of it will not appear in ChatGPT search answers, though it can still show as a navigational link. GPTBot is a separate crawler for content that may be used to train models. ChatGPT-User handles actions a user starts in ChatGPT, such as opening a page. OpenAI says the OAI-SearchBot and GPTBot settings are independent of each other, and that robots.txt rules may not apply to ChatGPT-User, because a person starts those visits.

Connect on LinkedInSagnik Bhattacharya

Google's page on AI features says a page must be indexed and eligible to appear in Search with a snippet; there are no additional technical requirements, and no special structured data or AI text file is needed. It recommends ordinary fundamentals: crawling allowed in robots.txt and by your CDN or host, internal links, a good page experience, important content as text, structured data that matches the visible text, and an up-to-date Business Profile. And it says that meeting every requirement still does not guarantee a page will be crawled, indexed or served.

Subscribe on YouTube@sagnikteaches

So the technical bar is low and the content bar is where practices differ. The tools tend to assemble an answer from pages that state a clear claim about a person or firm, and from third-party pages that corroborate it. The five jobs below build both halves.

Job one: let the right crawlers in

Open your robots.txt (yoursite.com/robots.txt) and look for rules that block OAI-SearchBot, or a blanket "Disallow: /" aimed at every bot. Many security plugins and some website builders block AI crawlers by default. Decide on purpose. If you want ChatGPT search to be able to cite your site, allow OAI-SearchBot. If you do not want training use of your content, you can disallow GPTBot separately; per OpenAI, the two settings are independent.

Also check your CDN or host: a bot-protection rule can block crawlers even when robots.txt allows them. OpenAI recommends allowing its published IP ranges too. If you use a managed host and cannot read these settings, ask the support team a single question: "Do you block OAI-SearchBot or Googlebot?" Allow up to a day after any change before testing, since OpenAI says robots.txt changes can take about 24 hours to take effect.

Job two: one page that says exactly who you are

An entity page is simply the page a system would read to answer "who is this consultant?". For most practices it is the About page, rewritten as a fact sheet rather than a story. It should state, in plain sentences near the top:

• Your name and the name of your practice, spelled the same way everywhere.
• The one-line description: "an independent consultant who helps [type of client] with [specific problem]".
• The services, each named the way a client would search for it.
• Who you do not work with. This sharpens the match more than any adjective.
• Credentials, memberships and years in practice, only the true ones.
• Where to verify you: links to your profile on a professional body, your company registration, and your main social profile.

Illustrative before and after for a consultant in operations: "I partner with ambitious leaders to drive transformation" says nothing a system can use. "I am an independent operations consultant for 20 to 200-person manufacturers. I fix scheduling and stock problems, typically in 6 to 12 weeks, and I do not do software selection" gives the tool a category, a client type, a time frame and an exclusion. When a prompt asks for "an operations consultant for a small manufacturer", the second page matches and the first does not.

Job three: case studies written to be quoted

Case studies are the most quotable asset you own. Keep them honest: use a client name only with written permission, otherwise a generic description. A format that works for both humans and AI tools:

Client type: 45-person food manufacturer (anonymised, with permission)
Problem: Stock-outs on three ingredients, 12 percent late orders
What I did: Rebuilt the reorder points, set a weekly review
Time: 8 weeks
Result: Late orders down to 4 percent within a quarter
What did not work: First attempt at a shared spreadsheet; replaced by
a simple weekly meeting
Date: March 2026

The figures above are an illustration, not a real engagement. Yours must be real and documented: do not publish an outcome you cannot show a client agreed to. The "what did not work" line matters; it reads as evidence, and it separates your page from the generic ones.

Give each case study its own page with a descriptive title ("Cutting late orders for a 45-person food manufacturer"), not a slide gallery on one page. Link each from the relevant service page. For help judging what buyers look for, the tutorial on vetting a consultant's case studies is written from the buyer's side and shows what they will read closely.

Job four: reviews and profiles on pages other than your own

AI systems lean on corroboration. A claim on your site is a claim; the same claim on a directory, a professional body page and a client's LinkedIn recommendation is evidence. List the places a buyer in your field would look and make sure each carries the same name, description and links:

1. Your professional body or association member directory.
2. Your company's public registration entry, if you are incorporated.
3. Google Business Profile, if you serve clients in person.
4. LinkedIn, with the same one-line description.
5. One or two review sites that your clients genuinely use.
6. Podcasts, guest articles or event pages where you are named, each linking back.

Ask for reviews that describe the work in words: "helped us fix X in Y weeks" is more useful than five stars. Do not buy or write fake reviews, and do not offer money for good ones: it breaks platform rules, can breach consumer law, and it is exactly the sort of thing automated systems and humans learn to discount. If you want a framework to evaluate paid services in this area, the tutorial on AI visibility providers lists what to ask.

Job five: structured data that matches the page

Google says no special markup is needed for AI features, so this is the lowest-priority job and a good one to do last. If you add it, the rule is that markup must match the visible text. A Person or ProfessionalService entry in JSON-LD gives search engines your name, description, URL and profile links in a standard form. Use only facts that appear on the page, validate with Google's Rich Results Test or the Schema Markup Validator, and put the same details in your visible footer.

A short illustrative example of the markup shape for a solo practice:

{
  "@context": "https://schema.org",
  "@type": "ProfessionalService",
  "name": "Example Operations Consulting",
  "description": "Independent operations consultant for 20 to 200-person manufacturers.",
  "url": "https://www.example.com/",
  "sameAs": [
    "https://www.linkedin.com/in/your-profile"
  ]
}

If you use a plugin or an AI assistant to generate markup, validate the result. The schema generator tutorial explains how to check output before launch, which is the step people skip.

The fixed prompt test you rerun every month

You cannot measure rank in an AI answer the way you measure a search position, because answers vary between runs, users and accounts. The honest approach is a fixed panel of prompts that you run on a schedule and log. Build ten prompts: some that name your category, some that describe a buyer's problem, and a few that name you directly.

Run each prompt in a logged-out or fresh session, in ChatGPT with search switched on, in Google's AI Mode, and in one other assistant. Run each twice, because answers differ between runs. Log the date, the tool, whether you were named, which sources were cited, and anything wrong. Keep it in a sheet. The brand prompts are the most revealing: if the answer misstates your services or credentials, you have a specific page or profile to fix, and you have found out before a client did.

A worked illustration: a consultant runs ten prompts twice in three tools, so 60 answers. In month one she is named in 3 of 60, with 2 of the 3 for the brand prompt she wrote herself. After rewriting the entity page, publishing two case studies and opening robots.txt, month two shows 7 of 60 and the category prompts now cite a case study page. That is illustrative, and it is not a promise: a change from 3 to 7 out of 60 sits inside the normal swing between runs, so she waits for the third month before drawing a conclusion. Treat anything under a dozen named answers as a direction, not a measurement.

If you want a tool to do this for you, some SEO suites track brand mentions in AI answers, and the Semrush and Ahrefs comparison for agencies covers two of them. For a single practice, a spreadsheet and thirty minutes a month is plenty.

Reading a bad answer: a brand prompt gone wrong

The brand prompt is where you learn the most, so here is an illustration of how to use it. Say you ask an assistant, "What does [your name] do?" and the reply says you are "a business coach who runs leadership retreats". You are an operations consultant. Do not argue with the tool; find where it got that. Look at the cited sources, and you may find an old conference page where you spoke on leadership, a directory entry written years ago, or a LinkedIn headline you never updated. Fix the source, not the symptom: change the profile, ask the event organiser to correct the bio, and make sure your own entity page states the correct description in the first two sentences.

Rerun the prompt after a week. If the wrong description persists, the tool may be leaning on a page you have not found; search for your name in quotes and check page two and three of results. This is detective work, and it is the best argument for the monthly test: you catch the problem when it costs nothing.

Three practices, three different emphases

An independent financial-controls consultant. Buyers ask problem prompts ("who can review our month-end close process?"), and trust matters because the work touches the books. Emphasis: credentials stated precisely, one or two anonymised case studies with a measurable result, and a service page named for the problem, such as "Month-end close review for owner-managed companies".

A strategy consultant for a narrow sector. Here the category prompt dominates ("strategy consultants for independent veterinary groups"). Emphasis: the sector named on every service page, content that shows you know its vocabulary, and mentions on sector association pages. A narrow claim is easier for a system to match than a broad one, so resist the urge to widen it.

A part-time consultant with a day job elsewhere. Fewer pages and fewer profiles, so corroboration is scarce. Emphasis: do the entity page and one case study properly, get the professional body listing right, and accept that the test results will move slowly. Three accurate sources beat ten inconsistent ones.

A service page structure that gives AI tools something to quote

Each service page can follow the same small template, which also reads well to humans:

H1: [Service name as a client would search it]
Opening paragraph: who it is for, what the problem is, what you deliver,
  in 50 words or fewer.
Who this suits: three bullet points with size and situation.
Who it does not suit: two bullet points.
What happens: 4 to 6 numbered steps with typical time.
Typical result: one sentence, linked to a case study page.
Questions: 4 real questions with 40 to 80 word answers.
Next step: how to book a call.

The opening paragraph and the question-and-answer block are the parts most likely to be lifted into an answer, so write them as standalone statements. "Month-end close review: for owner-managed companies with 10 to 80 staff whose close takes more than ten working days. I map the steps, find the delays and agree a shorter calendar" works whether it is quoted alone or read in context. Avoid pronouns that need the previous paragraph to make sense.

The question block should answer questions your clients really ask in the first call: how long, what do you need from us, what does it cost to start (a range or "fixed fee after a free scoping call"), what happens if it does not work. Do not invent prices; if you do not publish fees, say how you quote.

When a result is noise, and when it is a signal

With a panel of ten prompts, run twice, across three tools, you have 60 answers a month. Natural variation between runs can move a count of named answers by a few either way, so set a threshold before you look. A practical rule: act on a change only if it persists across two consecutive months or if a brand prompt returns a factual error. Record each change you make to the site, with the date, in the same sheet, so you can see which change preceded which movement. Without that record you will credit whichever change you remember best.

Finally, keep the test honest. Do not tune a prompt after seeing a poor answer to make it return a better one, and do not count an answer where you are named only because the prompt contained your name unless it is a brand prompt. The category and problem prompts are the ones that reflect new enquiries.

What not to promise yourself, or a client

No one can guarantee that an AI tool names you. The answers depend on the prompt, the account, the date and the sources available; they change when models change. Anyone who sells a guaranteed mention is selling something nobody controls. What you can control: access, clarity, evidence and corroboration, plus an honest log of results. If a provider's pitch includes "we will get you into ChatGPT", ask them what they would do, in steps, and how they would prove it, and compare it to the five jobs here.

Also be wary of one specific trap: asking an AI tool to write your case studies or reviews. A generated case study with invented figures is worse than none. If a system quotes a fabricated number from your page, you have created a false claim in the public record under your name.

A two-week plan for a sole practitioner

1. Day 1: read robots.txt, ask the host about bot blocking, write down the findings.
2. Days 2 to 3: run the first ten-prompt test and log it. This is the baseline.
3. Days 4 to 6: rewrite the About page as an entity page and align your LinkedIn description to it.
4. Days 7 to 10: write two case studies, get written client approval, publish each on its own page.
5. Days 11 to 12: check the directories and professional body listing; fix inconsistencies; ask two clients for descriptive reviews.
6. Days 13 to 14: add and validate markup; set a monthly reminder for the test.

The whole thing takes about 15 to 20 hours spread across the fortnight, mostly writing. If you decide to hire help for part of it, use the questions in the providers tutorial first, and check that what you are buying matches the five jobs. For a trade or local-service view of the same signals, the plumbers' tutorial shows how they change when customers search by location and urgency.

Questions consultants ask about AI search

Can I pay to be recommended by ChatGPT?

Not inside the answer. OpenAI says ads on ChatGPT's Free and Go plans are clearly labelled, kept separate from the answer and do not influence what ChatGPT says. Treat any provider who promises a guaranteed mention with suspicion and ask what exactly they are selling.

Do I need an llms.txt file?

Google's guidance says AI features need no special files or markup. Other tools may read such a file, but I would not spend time on it before fixing the page content and the profiles AI tools actually cite. Add one later if you want, as a minor extra.

Is schema markup worth adding?

Google says no special structured data is needed for AI features, but accurate markup that matches the visible page text helps search engines understand who you are. Add Person and ProfessionalService markup where it is true, and validate it. Do not expect it to move you into answers alone.

How long before changes show up?

OpenAI says robots.txt changes can take about 24 hours to reach its search systems. Content changes depend on when pages are crawled and when profiles elsewhere update, so test monthly rather than daily and record the date of each change.

Further reads

• Can Estate Agents Be Found in AI Search? What to Fix Now — Another professional-services take on being found by AI.
• AI Visibility Tracking for B2B: Affordable Tools and Free Checks — Cheap and free ways to track AI mentions.
• Best AI Brand Visibility Software: Is AI Mentioning Your Business? — Software that checks whether AI mentions your business.
• Consulting Pricing Models After AI: Fixed Fee, Retainer or Value? — Once found, how to price what you sell.
• AI in the Consulting Industry: Package a Fixed-Scope Diagnostic to Sell — A fixed-scope offer to put on your service page.
• SEO for AI Search: A Client Audit Checklist for Agencies — A seven-part agency audit checklist: crawler access, indexability, answer clarity, entity facts and citations, tied to Google's own AI-search guidance.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI's crawler documentation (OAI-SearchBot, GPTBot, ChatGPT-User), OpenAI's post on its approach to advertising in ChatGPT, and Google Search Central's 'AI features and your website' page, all read 4 November 2026.