Why isn't my SaaS appearing in ChatGPT? A practical checklist

If your SaaS is missing from a ChatGPT answer, begin with the question and the evidence behind that answer. One missing mention cannot tell you whether your website is inaccessible, your product is a poor fit for the question, or the answer simply selected other options.

The useful goal is to identify a specific issue you can investigate. This checklist separates access, relevance and product information so you can make a defensible next decision.

Start here: what did you actually observe?

What you observed	What it establishes	Your next check
The answer did not use web search	It did not test current search-backed visibility	Save it separately; record whether a later check uses search
Your brand was absent from a search-backed answer	No mention in this saved answer	Inspect the exact buying question and cited sources
Your brand was named, but its site was not cited	A mention, without a citation to your site	Check whether the description is accurate
Your page was cited, but your product was not recommended	A citation, without a recommendation	Read what the citation actually supports
A public product page returns a login wall, error or challenge	An access obstacle for that request	Ask your developer to check that page and crawler handling
The check failed or returned no usable answer	Missing evidence	Record unavailable; do not count it as an absent mention

A page opening in your browser does not, by itself, prove a search crawler can fetch it.

1. Confirm what you actually checked

Record the exact question, date and whether the answer used web search. A conversation that asks ChatGPT to recall products without searching is a different observation from a search-backed answer with current source links.

Also record whether you had already described your company or your preferences in the conversation. For a cleaner category check, begin a new conversation without first teaching the assistant about your brand. Record relevant memory and location settings rather than assuming a fresh chat removes every source of personalization. OpenAI explains that location and, when enabled, relevant memories can influence searches. How ChatGPT searches the web

Record brand mentions and source citations separately. Your SaaS can be named without its website being cited, and a page can be cited without the answer recommending the company.

Separate these two questions:

• “What does AcmeDesk do?” checks recognition and description of a named product.
• “Which shared inbox tools suit a five-person SaaS support team?” checks discovery within a buying situation.

Both can be useful. An answer to the first does not establish that buyers encounter you through the second.

2. Check search access before rewriting your content

Ask your developer to check the important public product pages, not just the homepage. Can an ordinary visitor reach them without logging in? Do they return useful content, or a challenge screen, error or empty shell?

For ChatGPT search, OpenAI identifies OAI-SearchBot as its search crawler. Its documentation distinguishes that crawler from GPTBot, which concerns potential model-training use. These settings are independent. Allowing search access does not require making the same choice about training. OpenAI's crawler documentation

Have the developer inspect the site's crawler rules and hosting or security-layer handling against that documentation. Keep intentional restrictions on private application and account pages. Public product information is the relevant target.

If you find an access problem, fix it and record the change. If access is already working, move on. Allowing a crawler creates an opportunity for discovery; it does not establish that your product will be selected for a particular answer.

3. Ask whether the buyer question fits your product

A broad category question may favor products with capabilities your SaaS does not offer. That makes it a weak diagnostic of your strongest market.

Write down three things your actual buyer brings to the decision:

1. The job they need to accomplish.
2. The type and size of their team.
3. A constraint that changes which products fit.

For a hypothetical shared inbox product, a relevant question could be: “Which support inbox tools let a small B2B SaaS team coordinate email replies without setting up a full contact center?”

Do not add an unusual combination of features solely because it guarantees that only your product qualifies. The question should sound like something a real prospective customer would ask before knowing your name.

4. Inspect the page that should answer the question

Open the public page that best supports your buying use case. Can a reader identify the product, its audience and its useful capabilities from the text?

For the shared inbox example, the page should explain who shares the inbox, how replies are assigned and what happens when two people respond. A heading such as “The future of collaboration” does little to answer those questions by itself.

Use a small evidence table to identify gaps:

Buyer needs to know	Public evidence to inspect
Does this fit our team?	A specific audience or use-case explanation
Can it solve our workflow?	A clear capability description and example
What will implementation involve?	Setup requirements and supported integrations
What would make it unsuitable?	Honest limitations and plan restrictions

Improve missing or ambiguous information because it helps the buyer evaluate your product. Treat any later change in AI answers as an observation to investigate, not proof that one edit caused a recommendation.

5. Read the sources attached to competing answers

When competitors appear, inspect the cited pages. They may be product documentation, reviews, comparison articles or other third-party material.

Look for the information those pages provide. Are they clearer about the use case? Do they answer a buying constraint that your own page omits? Is the answer relying on an outdated description that you can document and correct at its source?

A citation does not reveal the platform's complete selection process. Use it to identify material worth reading, not as a blueprint that promises the same result if copied. Do not manufacture reviews or unsupported claims to imitate another company's apparent authority.

Worked example: an absent mention with a useful next step

Fictional example, not an AnswerProbe test or customer result. AcmeDesk is an invented shared inbox product for small SaaS support teams.

The founder saves a search-backed answer to: “Which shared inbox tools help a five-person SaaS team assign email replies and avoid duplicate responses?” The answer names three other products and links to their documentation. AcmeDesk is absent. A separate question naming AcmeDesk produces a broadly accurate description.

The founder opens the cited documentation and AcmeDesk's public page. The competitors explain reply assignment and collision warnings. AcmeDesk's page says “Work together effortlessly” but never explains either capability. Its internal product documentation confirms both exist.

Evidence saved	Defensible conclusion
AcmeDesk absent from the buying answer	It was not mentioned in this observation
Accurate answer to the named-brand question	Recognition appeared in a separate observation
Public page omits two verified capabilities	There is a specific information gap worth fixing

Still unknown: why the answer selected those competitors. The missing explanation does not prove the cause of AcmeDesk's absence, and recognition does not prove recommendation eligibility.

One action: add a factual assignment example, explain collision warnings and state which plans include them. The product-page checklist provides a review structure. Keep the original question and saved answer. Compare later search-backed observations with the same wording and recorded settings, including any continued absences. A later mention would be an encouraging observation, not proof that the page edit caused it.

6. Choose one change and keep a comparison record

Your first action should follow the issue you found: correct an access failure, clarify a product capability, document a relevant integration, or investigate a misleading third-party description.

Save the question, answer, sources and date before making the change. Repeat a comparable check later and record unavailable results separately. An unavailable observation is missing coverage, not zero visibility. Avoid rerunning until you obtain a favorable screenshot and presenting only that answer.

Copy this diagnostic record

Field	Your record
Product and public page
Buyer job, team and constraint
Exact question
Engine/product, date, time and timezone
Web search used	Yes / no / unknown
Relevant memory, location and conversation context
Saved answer and cited URLs
Result	Available / unavailable
Brand mentioned	Yes / no / not assessable
Own website cited	Yes / no / not assessable
Specific product recommended	Yes / no / not assessable
Verified access issue or information gap
What remains unknown
One change and date applied
Comparison date and any changed conditions

Keep unavailable checks and unsuccessful comparisons in the record. Do not replace the baseline with your most flattering answer. Our explanation of why AI answers change helps separate changed conditions from a comparable observation.

Our guide to tracking company mentions and citations in AI answers explains how to preserve the questions, answers and sources for a comparable record over time.

AnswerProbe's AI visibility checker page explains its free Brand Probe: one approved customer need, three reviewed questions and three answer engines, for nine planned observations. You enter your brand information manually. A verified account, confirmed eligibility and your explicit approval to execute are required. Available results preserve evidence you can inspect. Coverage can be incomplete, and reviewed interpretation depends on successful processing and available capacity; the result can be labelled limited analysis. These controlled observations are not a universal ranking of your business.

The best immediate outcome is a clearer diagnosis and a useful next action. A missing mention becomes actionable when you can connect it to evidence, rather than treating it as a universal verdict on your business.

Frequently asked questions

Why can ChatGPT describe my SaaS but leave it out of a recommendation?

A question that names your product tests recognition. A question about solving a buyer's problem tests discovery and fit. Save both answers, but use realistic buyer questions to investigate whether your SaaS appears before the buyer knows its name.

Do I need to allow GPTBot to appear in ChatGPT search?

No. OpenAI documents OAI-SearchBot for search and GPTBot for potential model-training use, with independent settings. Ask your developer to check access for the search crawler, including the hosting or security layer. Search access does not guarantee placement.

Will updating my website make ChatGPT recommend my product?

There is no guaranteed edit or timetable. Correct a documented access problem or a specific information gap, then compare later observations under similar conditions. A changed answer alone does not prove that your edit caused it.

Can a free AnswerProbe check explain exactly why my SaaS is missing?

No. It checks a small, approved set of questions and preserves available answer evidence. It cannot reveal an answer engine's internal selection process. Coverage can be incomplete, and reviewed interpretation is not guaranteed; a result may be labelled limited analysis. Inspect the evidence, keep unavailable results separate and use any reviewed interpretation as a starting point for investigation.