If ChatGPT does not recommend your business, check the question asked, the sources shown, your site's search access, and whether your pages explain the buyer's actual need. One missing recommendation does not establish why you were absent. The useful outcome is a documented list of things you can verify and improve.
Seeing a competitor named when your company is absent can make the next decision feel urgent. You may be ready to rewrite the homepage, buy a visibility tool, or publish a stack of articles. Before spending that money, save the answer and work out what you actually know.
A useful review should end with a task someone can own: investigate an access problem, correct a business detail, explain a service more clearly, or collect more observations. An unsupported theory about why the system prefers a competitor is a poor brief for your web team.
Keep the answer that started the question
OpenAI says ChatGPT search can use location and relevant saved memories, and search placement is not guaranteed. See its explanation of ChatGPT web search. That makes the circumstances of your test worth recording.
- Save the complete question, answer, and date, including the sources shown.
- Record whether web search was used and which product or mode you tested.
- Note the location in the question and relevant account or conversation context.
- Keep the businesses named and the actual words used to describe them.
Then read the question as a buyer would. A request for a roofing company that repairs tile roofs is different from a request for a contractor to replace an asphalt roof. Keep your test tied to a service and area you actually serve.
Use separate questions to check whether the answer describes your company accurately and whether it recommends your company without being prompted by name. Adding your name changes the exercise. Asking why you were excluded may suggest things to investigate, but require supporting evidence before treating that explanation as the cause.
Ask your web team to check search access
OpenAI distinguishes OAI-SearchBot, which supports search discovery, from GPTBot, which concerns possible model training. Their settings are independent. Its guidance recommends allowing OAI-SearchBot and OpenAI's published searchbot IP ranges. The crawler documentation explains those controls.
OpenAI's publisher guidance says access by OAI-SearchBot matters for including site content in summaries and snippets. Its publisher FAQ describes that scope. Review your actual settings before deciding that a missing answer is a writing problem.
Ask the person responsible for your site to inspect the relevant pages, robots.txt rules, and hosting or security controls. Give them the URLs you want reviewed. Ask whether an intended public page loads successfully and whether an unintended rule blocks the search crawler. Keep the findings with the date and evidence.
Opening a page yourself is a useful first check. It does not answer every question about crawler access. Similarly, a recorded bot request establishes a request; it does not establish that your business was recommended. If the web team finds no access issue, continue the review without inventing one.
Does the page describe the work the buyer needs?
Imagine a manufacturer that accepts small production runs but whose website says only that it provides metal fabrication. This is a hypothetical example. A buyer looking for a supplier for a limited run would still need to know whether that work fits the company, what information to submit, and which constraints affect the discussion.
If those answers exist only in a salesperson's head, start there. Ask the business to confirm its scope, customer fit, service area, and limits. Find out which details repeatedly have to be explained before a useful inquiry can move forward.
Read the page without filling in the gaps from your own knowledge. Could a prospective customer explain what you do, whether you fit the job, and what happens after contact? If the answer depends on assumptions, write the missing explanation using facts the business can support.
This is the practical work of finding unanswered buyer questions. A clear answer helps someone assess the business. Treat it as an improvement to the page, with its search impact still to be observed.
Check what supports the explanation as well. A claim about capability is more useful when the reader can inspect relevant work, a documented process, or another appropriate piece of proof. The proof your website needs should follow the claim you are asking a buyer to believe.
Read the sources behind the recommendation
OpenAI cautions that search results and citations can be incomplete, outdated, or incorrect. Open the linked sources and inspect them before building a plan around what an answer says.
For each relevant source, ask what information it contains about the business, whether that information is current, and whether it supports the description in the answer. A source might be the company's own page, a directory entry, or another publication. Record what you can actually see.
If your business appears with an old name or an incorrect service description, identify where the error is published and who can correct it. If a competitor's page explains a buyer concern that yours leaves unanswered, record that concern. Neither observation tells you the weight the system gave that page.
Use the comparison to discover useful information to add or correct. Copying a competitor's wording would leave you with their explanation of their business. Your page needs the facts, conditions, and examples that belong to yours.
Turn findings into a short work list
I would assign work in the order the evidence supports. A confirmed access failure deserves a technical task. An inaccurate public detail needs correction. An unanswered customer question needs an approved explanation. A page that already answers the question may need no immediate rewrite.
Write each task so another person can complete it: name the affected page, describe the issue, identify the information needed, and state who will check the result. Keep a record of the original version. That makes it possible to explain what changed when you review the page again.
If several pages appear to need the same explanation, decide which page should own it before commissioning new articles. Reviewing whether to update a page or create new content keeps the work focused and gives related pages a sensible place to point readers.
Repeat the check and keep the result in proportion
Choose a recurring check your team can maintain. Repeat the useful questions under comparable conditions, save the complete answers, and note any changes to the site or test setup. Keep questions that produce an unfavorable result in the record too.
Read the result carefully. An answer can mention a business without recommending it, and a source link can support an explanation without endorsing the company. Measuring mentions, citations, visits, and leads helps you keep those observations separate.
The broader guide to getting your business found in AI search connects this investigation to the rest of the content work. Start with the issue you can substantiate, make a useful improvement, and keep enough evidence to assess what happens next.
Discuss what your website needs next
If your business is missing from the answers you are checking, book a 30-minute call with Hudson South to discuss your content priorities and the questions your website needs to answer.
Book a 30-minute callSources
- Searching the web with ChatGPT. OpenAI. Read 2026-10-08.
- Overview of OpenAI Crawlers. OpenAI. Read 2026-10-08.
- Publishers and Developers - FAQ. OpenAI. Read 2026-10-08.
