Why your business doesn't appear in ChatGPT (and how to get in)

Why your business doesn't appear in ChatGPT

If ChatGPT cites your competitors and never you, three reasons explain it in almost every case. First, your external authority is insufficient: ChatGPT only knows you if credible third-party sources talk about you, and you don't have enough of them. Second, your content is not answer-first: your pages describe who you are instead of directly answering the questions your customers ask, so the model finds nothing to extract. Third, structured signals are missing: without schema.org markup and without machine-readable content, you are invisible to the engine that assembles the answer. The first two causes carry the most weight. The good news: they are largely within your control. We explain why, with the figures, then what depends on you and what does not.

Reason 1: your external authority is insufficient

ChatGPT does not read the web live, it relies on a search index to answer current questions, and that index is Bing's. The Seer Interactive study (analysis published in 2024 on hundreds of queries) measured that the answers cited by ChatGPT come about 87 % from the top 10 Bing results. In plain terms: if you are not in the top ten Bing results on your customers' queries, your probability of being cited by ChatGPT collapses. The Bing ranking itself depends on your perceived authority, therefore on the links and mentions other sites make about you.

This weight of third-party sources also appears in academic research. A study published on arXiv in 2025 (reference 2509.08919) analysed the origin of AI answer engine citations by sector: between 73 % and 92 % of cited sources are third-party content, what marketing calls earned media (press, specialised directories, comparison sites, articles by other players), and not your own site. In other words, your site alone is almost never enough. The model seeks external corroboration before naming you. A brand with a thin digital footprint, rarely cited elsewhere, is structurally ignored by LLMs, not out of malice but because no signal brings it up.

Reason 2: your content is not answer-first

An LLM cites what it can extract in one go: a clear, self-contained answer, phrased as an answer. Most company sites do the opposite. They open with a commercial hook, tell their story, then bury the useful information in the middle of a paragraph. The model, meanwhile, looks for a block that directly answers the question asked. If it doesn't find it on your site, it takes it elsewhere.

The answer-first format means answering first, developing afterwards. Each page handles one precise question, answers it in a complete sentence from the first line, then goes into detail. This is the number one citation factor on Perplexity, and the principle holds for every answer engine. We detail which sources the models go looking for in our analysis which sources ChatGPT cites. Content written to seduce a rushed human visitor is not content written to be cited by a machine. These are two different exercises, and most businesses have only done the first.

Reason 3: structured signals are missing

The third obstacle is technical. Without schema.org markup, without a clean heading hierarchy, without content accessible to AI indexing robots, the engine struggles to understand what you say and to reuse it confidently. Article and FAQPage markup gives the model an explicit reading of your questions and answers. A robots.txt file that allows GPTBot, ClaudeBot and PerplexityBot ensures these robots can actually read your pages.

Be careful, this signal is necessary but not sufficient. Schema does not create authority: it makes readable a content that must already be good and corroborated elsewhere. That is why we rank it third. Many agencies sell markup as the miracle cure. It is not. Schema speeds up and stabilises machine reading, it replaces neither substance nor external authority.

What is within your control

The actionable share is large, and that is where your visibility is decided.

• ·Answer-first format on your site. Rewrite your pages so they directly answer your customers' questions. It is the fastest and most cost-effective lever.
• ·Schema.org markup. Add Article, FAQPage, Organization, and a robots.txt that allows the AI robots. Technical work, lasting effect.
• ·Substance content. Produce dense, sourced answers on the real questions of your market. This is what makes you extractable and corroborable.
• ·Trigger external mentions. Local press, sector directories, partnerships, interviews, contributions. Each credible third-party mention raises your perceived authority, therefore your Bing ranking, therefore your probability of being cited.

Our AI visibility diagnosis tests exactly these signals and tells you where you stand, for free.

What is not within your control

Let us be honest about the limits.

• ·Training data already frozen. What the model learned during training is set until its next version. You do not rewrite the past. You work the live search layer, which does move.
• ·Your competitors' historical authority. A player cited for years starts with an inertia lead. You do not erase it in a month, you catch up gradually.
• ·Model publishers' internal decisions. OpenAI, Anthropic and the others change their sources, their indexing partnerships and their rules. No one outside these companies controls that.

Our job is to maximise what depends on you and to measure the result, without ever claiming to steer what does not depend on you.

How long before being cited

The honest answer: it depends on your starting point. If you already have an authority base (an indexed site, existing external mentions, a sector presence), the answer-first and schema compliance produces observable effects in 2 to 6 weeks, the time for engines to reindex and answers to update. For a new domain, with no history or third-party mentions, expect rather 4 to 6 months: you first have to build the missing external authority, and that cannot be bought in a week.

These ranges are field observations, not promises. We work a probability of being cited, we sell no citation guarantee. No one can guarantee that a model will name you on a given query, because no one outside the model publishers controls the final output. What we guarantee is to lay down all the signals that tip the probability in your favour, and to measure it every month with dated screenshots. For the specific case of Perplexity, see our guide getting cited by Perplexity.

Frequently asked questions

Why does ChatGPT cite my competitors and not me?

Because they combine three advantages you don't have yet: more external authority (therefore a better Bing ranking, from which ChatGPT draws about 87 % of its citations according to Seer Interactive, 2024), more answer-first content the model can extract directly, and often markup that makes their pages machine-readable. Citation is not chance, it is the result of signals they have and that you can build in turn.

How long to appear in ChatGPT?

From 2 to 6 weeks if you already have an authority base (indexed site, external mentions, sector presence), the time for engines to reindex your corrected pages. From 4 to 6 months for a new domain with no history, because you first have to build the missing external authority. These are field observations, not guarantees: a model's final output cannot be steered.

Is schema.org enough to be cited?

No. Schema.org markup makes your content readable and reliable for the machine, but it does not create authority and does not replace substance content. It is a necessary signal, not a sufficient one. A page that is marked up but empty of useful answers and without external corroboration will not be cited. Schema speeds up machine reading of content that must already be good and recognised elsewhere.

Can being cited by ChatGPT be guaranteed?

No, and beware of anyone who promises it. A model's final output depends on its training data, its search layer and OpenAI's internal decisions, three elements no provider controls. What can be worked on is the probability of being cited: laying down the authority, answer-first format and markup signals that tip the balance, then measuring the result over time.