How to get your brand mentioned in ChatGPT

You built a product that deserves to make the shortlist. Ten minutes before a sales call, a buyer asks ChatGPT who they should consider. Your company is missing, described badly, or flattened into a list of interchangeable vendors.

That buyer may never see the product page you spent weeks refining.

Technical founders feel this gap sharply. You solved a real problem, found customers, and built something that works. Yet a thinner competitor can look safer because its story is easier to find and repeat. Good companies should not lose simply because the market cannot understand them.

We understand why this work gets postponed. Shipping, hiring, fundraising, and customer work take priority. At Unfungible, we operate Founder Funnel for founders who know their product far better than they explain it publicly. Our job is to install the content infrastructure that turns what the founder already knows into evidence a buyer can find.

Your buyer may research you before the call

A prospect once joined a sales call with me after researching Founder Funnel in ChatGPT. I asked what he knew about us. He said, "I just asked ChatGPT to look you guys up before this call." He had not started with our homepage, my YouTube channel, or my social profiles. He had asked an assistant whether the conversation was worth his time.^[5]

His decision had already started forming before we spoke.

I later caught myself doing the same thing. A friend referred me to a law firm in the UAE. I gave the firm's name to Claude and asked what it thought. Claude summarized the firm and surfaced five alternatives. I contacted the referral and the alternatives, then hired one of the competitors because its product, customer service, response time, and pricing fit me better.^[5]

These are personal stories, not market statistics. They show what is at stake. An assistant can enter the buying journey before the company's site does.

If the assistant omits you, gets your category wrong, or cannot explain why you fit, the buyer has to do extra work to understand you. Some will keep moving. The cost is not an abstract "AI visibility score." It is a weaker shortlist, a sales call spent correcting basic facts, or a qualified buyer who never arrives.

First, define the outcome you want

"Mentioned in ChatGPT" can mean four different things. Track them separately.

A mention is not an endorsement. A citation can support a general claim without recommending you. A recommendation may name you without linking to your site. Sometimes the most useful early win is less glamorous: the assistant finally describes your company correctly.

These are measurement categories for your team. They are not official OpenAI ranking factors.

Search eligibility is different from model training

In the source video, I described AI as scanning the web and forming a composite picture from websites, articles, YouTube videos, social posts, podcasts, reviews, and Reddit threads.^[5] That captures the marketing problem, but the technical reality needs more care.

OpenAI documents separate controls for search and model training. OAI-SearchBot helps surface websites in ChatGPT search features. GPTBot crawls content that may be used to train OpenAI's generative models. A site owner can allow one and disallow the other.^[6]

If you want to become eligible for ChatGPT Search, inspect OAI-SearchBot. Do not treat the GPTBot setting as the same decision. OpenAI says sites that opt out of OAI-SearchBot will not appear in ChatGPT search answers, though they may still appear as navigational links.^[6]

ChatGPT can search the web when a question needs current information, and a search-backed answer may include citations. OpenAI warns that search results and citations can be incomplete, old, or wrong. It also says placement is not guaranteed.^[1]

Aim for eligibility and stronger evidence. Nobody outside OpenAI can promise control.

A four-part plan for earning consideration

Start with access. Then fix what your pages say, add outside evidence, and measure what the engines return.

1. Make the site eligible for live search

A healthy page in your browser can still return a 403 response to a crawler.

OpenAI recommends allowing OAI-SearchBot in robots.txt and requests from its published IP ranges. Its crawler documentation lists the current user agent and IP source.^[6]

• Check that robots.txt does not block OAI-SearchBot from the pages you want found.
• Confirm important pages return a successful response without a login.
• Test whether your CDN, firewall, or bot protection rejects the crawler.
• Keep important facts in the returned HTML.
• Use canonicals to identify the page you want treated as the source.
• Make the page discoverable through internal links and the sitemap.

Google takes a similar position on its own AI features. A page must be indexed and eligible to appear in Google Search with a snippet before it can become a supporting link in AI Overviews or AI Mode. Google says you do not need special AI schema, a special AI file, or extra technical requirements.^[3]

Technical SEO is still part of the job. Crawlers need access, and useful pages need to be discoverable.

2. Make your owned pages easy to understand

Technical founders often know too much about their product to explain it simply. A product conversation turns into architecture, features, and edge cases. The buyer leaves without five basic answers.

• Who is this for?
• Which costly problem does it solve?
• What does the company provide?
• What proof can the buyer check?
• When is it a fit, and when is it not?

In the video, I described how founders answer a positioning question by listing capabilities that competitors also claim. The buyer ends up with a basket of similar options and little reason to choose one.^[5]

Features matter. They need context.

An AI infrastructure founder should name the team, workload, constraints, and tradeoffs. A Web3 founder should state the network, custody model, user, risk boundary, and evidence behind security claims. A SaaS founder should name the workflow, buyer, implementation needs, and the conditions where another option fits better.

Run the shortlist evidence check

Give someone outside your company five minutes with your public pages. Can they find these five things?

1. The buyer. Name the role, company stage, and situation you serve. "For modern teams" is fog. "For security leads at AI companies shipping customer-facing agents" gives the reader a boundary.
2. The costly problem. Say what breaks, what it delays, or which decision it blocks. Use the buyer's language.
3. What you provide. Use one stable sentence across the homepage, article library, company profiles, and founder bios.
4. Proof they can inspect. Link to the original case, demonstration, documentation, methodology, or customer statement. Keep the scope honest.
5. Fit constraints. Say who should not buy. Name public technical requirements, service limits, stage requirements, and better alternatives for a different buyer.

If a stranger cannot build a fair shortlist from your own pages, an answer engine has little clean material to work with.

3. Publish pages around buyer questions

Your homepage cannot answer every decision that comes up during evaluation. Give each important buyer question one strong page. Answer it near the top, then add the method, evidence, examples, limits, and next action. If a page already owns the question, improve it instead of publishing a near duplicate.

Useful questions for a technical company might include:

• How should an AI company explain model accuracy without overstating it?
• What should a buyer check before adopting a Web3 custody product?
• When should a SaaS team build an internal tool instead of buying one?
• How does your product compare with the default way buyers solve the problem today?

A strong answer page responds directly, defines the decision, shows the method, cites the original evidence, adds the founder's judgment, and gives the reader a sensible next step.

This is where founder-led content helps. The founder can explain why the team accepted a tradeoff, what customers misunderstand, and where common advice breaks. Generic marketing copy usually strips out those details.

Use videos and calls as source material, not finished articles. A transcript carries spoken repetition and claims that may need checking. Keep the useful story. Verify the facts. Rebuild the piece around the buyer's question.

If your owned pages still read like product notes, use the founder-led marketing guide for technical founders as a lower-friction starting point.

4. Earn independent corroboration without faking it

Your site explains what you say about yourself. Buyers also want to know what customers, partners, publishers, and practitioners say.

OpenAI does not publish a rule that a set number of third-party mentions will trigger a recommendation. Treat outside coverage as useful evidence for buyers and another chance to be discovered, not as a guaranteed ranking lever.

Rand Fishkin recommends finding pages that already discuss the category and earning relevant inclusion there. He also warns that estimates about which sources a model uses are indicative, not exact.^[4]

Depending on the company, credible corroboration could include:

• a complete profile on a review platform that fits the product category;
• honest reviews from real customers;
• an independent comparison that describes the product accurately;
• a partner's integration page;
• a technical podcast or YouTube demonstration;
• a transparent community answer that discloses the founder's connection;
• original research or documentation that another writer chooses to cite.

Ask one question before pursuing any placement: would this page still help the buyer if it never changed an AI answer? If not, skip it.

Fake reviews, planted Reddit threads, self-serving "best tools" pages, and unsupported journalist pitches make the buyer's research worse.

Measure every answer engine on its own

ChatGPT Search, Google AI Mode, Perplexity, Claude, Gemini, and Copilot do not return the same answers or expose sources in the same way. A single screenshot only proves what happened in that run.

Build a fixed set of prompts around real buying situations. Include informational, commercial, comparison, and social-proof questions. Keep your company name out of prompts meant to test discovery.

Google includes traffic from AI features in Search Console's Web search type rather than a separate AI report.^[3] Keep your manual citation snapshots separate from Search Console and conversion data.

OpenAI says ChatGPT Search citations can be incomplete, outdated, or wrong.^[1] An incorrect description is not a win, even when your name appears.

Repeat the same prompt set. Compare trends across several runs. Do not turn one answer into a claim that you "rank number one in ChatGPT."

A practical 30-day checklist

The first month has a modest goal: remove preventable blocks, make the company understandable, publish one useful answer, and establish a baseline.

Days 1 to 5: establish the baseline

• Choose ten prompts that reflect real buying questions.
• Run them in every answer engine you can access.
• Record mentions, citations, recommendations, accuracy, competitors, and sources.
• Save the date, model, exact prompt, and whether search ran.
• Mark any engine you could not check.

Days 6 to 10: verify access

• Review robots.txt for OAI-SearchBot and Googlebot.
• Check OpenAI's current crawler documentation and IP ranges.^[6]
• Test your CDN and bot protection.
• Confirm important facts appear without authentication or extra interaction.
• Check canonicals, internal links, sitemap entries, and indexability.

Days 11 to 15: clean up company facts

• Write one stable sentence that names the company, buyer, problem, and offer.
• Use consistent company and product names on owned profiles.
• Put the offer, proof, and fit constraints in crawlable text.
• Remove claims you cannot support.
• Run the shortlist evidence check with someone outside the company.

Days 16 to 21: publish one buyer-answer page

• Choose a question that appears in sales calls and fits the offer.
• Give it one canonical page and answer it in the opening section.
• Cite primary documentation and original evidence.
• Add the founder's judgment, a demonstration, or a confirmed first-person story.
• Link the page from the relevant pillar, homepage, or documentation.

Days 22 to 26: strengthen outside evidence

• Inspect which independent pages your approved prompts cite.
• Correct inaccurate profiles you control.
• Ask real customers for honest reviews on a fitting platform.
• Find one comparison, integration, podcast, or publisher page where your evidence would help.
• Prepare any outreach for human approval. Do not automate community activity.

Days 27 to 30: rerun and decide

• Run the same prompt set again and compare each engine with its own baseline.
• Check description accuracy.
• Separate mentions, linked citations, and recommendations.
• Review referral traffic and qualified conversations without forcing attribution.
• Choose the next fix based on what you found.

Thirty days may pass without a visible change. You will still know whether the site is accessible, whether the company is described clearly, and where the evidence remains thin.

The cost of leaving this unresolved

A weak public record does not make your product worse. It makes the product harder to evaluate.

That creates ordinary, expensive friction. Sales calls begin with category corrections. Buyers compare you on price because they cannot see the difference. Partners repeat an old description. An assistant cites a page you never reviewed. A qualified buyer leaves the shortlist before your team gets a chance to help.

You do not need to chase every new AI feature. You need a reliable body of public evidence that survives changes in the interface.

Build the evidence before the next buyer asks

No consultant can guarantee that ChatGPT will mention your company. OpenAI says placement is not guaranteed.^[1] Google also says that meeting its requirements does not guarantee indexing or inclusion.^[3]

You can control whether crawlers reach your pages, whether those pages explain the company, whether the proof holds up, whether outside sources describe you accurately, and whether your team measures the right outcome.

A buyer asks an assistant about your category and gets a fair, accurate picture. They understand who you serve, what you solve, why your approach is different, and which evidence supports it. When they reach the sales call, your team can discuss fit instead of repairing confusion.