How a SaaS company gets mentioned in ChatGPT answers

Two ways ChatGPT knows about you

ChatGPT can answer from what the model learned in training, or it can search the web and cite pages. Many questions about software use search, especially recommendation and comparison questions.

• Training data. Built from large amounts of public text up to a cutoff date. You influence it slowly, by being described consistently in many places over time.
• Search. ChatGPT fetches pages at the time of the question. You influence it by being crawlable, by having a page that answers the question, and by appearing in sources that rank for it.

This guide covers the practical steps for one assistant. How assistants choose sources in general is covered in how AI assistants choose which sources to cite, and which formats tend to be cited in what content AI assistants actually cite.

Why your SaaS isn't showing up in ChatGPT recommendations

Usually for one of four reasons: ChatGPT cannot reach your pages, nothing you have published answers the question the buyer asked, the web describes your company inconsistently or out of date, or the other sources that rank for the question do not mention you. Each one can be checked, and the steps below cover them in order.

The cause is often mundane. On 3 October 2026, two days after this site changed its positioning, OwnedSignal ran a dated baseline on itself. An automated research agent put brand, category and buyer questions to two search tools: a web-search API that writes an AI summary of its results, and Exa. ChatGPT itself was not tested that day. The run found three problems:

• A name collision. Asked "What is OwnedSignal?", the web-search API's summary described a class with the same name in a Scala programming library instead of the company, on both runs.
• Old copies. Exa's index still held the previous version of the homepage and a retired page, so the old positioning was what it could retrieve.
• People merged. In one of two runs, an AI-written search summary combined three different people named Alex Iliescu into one biography.

These are observations from search tools, not ChatGPT answers. They show the kind of problem a buyer's assistant can run into when the web describes a company inconsistently, and they are worth checking in ChatGPT directly.

None of these is fixed by an AEO trick. They are fixed by consistent descriptions, current third-party profiles and time for crawlers to revisit, which is what the steps below address.

Step 1: make sure ChatGPT can reach your site

OpenAI documents its crawlers. OAI-SearchBot is used for ChatGPT's search features, and OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. GPTBot is a separate crawler used for model training. You can block training and still allow search.

Check your robots.txt and any firewall or bot-protection rules. Some security settings block these crawlers without anyone noticing.

Step 2: answer the questions buyers actually ask

ChatGPT searches for the question it was given, often rephrased into several searches. If your site only has a homepage, a features page and a pricing page, there may be nothing for it to cite on "how do I choose a vendor for X" or "what is the difference between A and B".

Useful pages for this are:

• Category explainers: what the problem is and the ways to solve it
• Comparison pages: your approach against named alternatives, written fairly
• "When not to use us" sections: these make comparisons more credible
• Specific FAQ pages for questions that come up in sales

Step 3: describe yourself consistently

If your site calls you an "AI data platform", LinkedIn says "workflow automation" and a directory says "analytics", a model has three competing descriptions. Choose one plain category description and use it everywhere. See entity clarity for search and AI.

The same applies to the founder. If the company site, the founder's profile and third-party bios describe the founder differently, assistants may leave them out or confuse them with someone else. Keeping the founder and company consistent is part of founder presence.

Step 4: be discussed elsewhere

AI answers frequently cite third-party pages: reviews, comparison articles, forums, video and interviews. Digiday reported, using Meltwater data, that YouTube was the most-cited platform across eight AI products in August 2026, while noting that a citation is often a recommendation to watch rather than evidence the video shaped the answer.

For a B2B company, the practical routes are customer reviews on the sites your buyers use, podcast and video interviews with the founder, partner pages and accurate entries in industry lists.

The mentions need to be real. Writing about its own AI features, Google says that "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem."

The shortlist is getting shorter

BrightEdge reported in September 2026 that ChatGPT named an average of 5.36 brands per answer to B2B technology prompts in March and 3.87 in September, a 27.9% decline. BrightEdge sells AI search tracking and measured mentions, not citations. If the direction holds, being sixth in a category now often means not being named.

The same report found that companies described as focused, single-purpose tools gained share at the consideration stage. That supports describing your company narrowly and precisely.

What you cannot control

• Which sources the model chooses on a given run
• How often the model is updated
• Whether a competitor is described more favourably by a third party

Answers also change between runs. Parse measured 161,023 back-to-back answers to identical prompts and found the top recommendation changed 43.6% of the time. Measure with repeated runs, not one screenshot. The method is in how to test whether AI assistants know your company.