Menu

Why Does ChatGPT Recommend Your Competitor? The Real Cost of Hidden Pricing and Closed Features

More and more B2B buyers no longer start their decision-making process on Google, but with a language model. If your SaaS is not included in the answer, the founder may think the AI “hallucinated.” The reality is more uncomfortable: the model simply did not see you — and you caused that yourself.

In short

ChatGPT, Claude, Gemini, and Perplexity work with what they can find about you publicly, structurally, and readably. Pricing hidden behind a “Request a quote” button and features that only exist behind a login result in the model recommending your competitor instead of you. Not because they are better — but because they are visible. The good news: this can be measured and improved.

The scene that now happens every week

Imagine a decision-maker at a 300-person company. They do not have time to browse ten websites manually. They open ChatGPT and type something like this: “Which [category] platform would you recommend for a 300-person company if SOC 2 compliance is important and the monthly budget is around €4,000?”

The model responds within seconds. Confidently. With three or four names, short explanations, pricing ranges, and even arguments for why those three are the right choices. This list becomes the buyer’s mental shortlist — before they open a single website.

And if your name is not on it, the deal never even starts for you. You did not lose because your demo was weak. You lost because the decision started in a room you never entered.

Not hallucination. A visibility gap.

When a founder sees that AI mentions their company incorrectly or not at all, the first reaction is almost always the same: “this AI is unreliable; it is hallucinating.” It is a comfortable explanation because it is not about us. But in most cases, it is misleading.

The language model is not lying about you. It simply fills the gap with what is visible. If there is not enough public, structured information about you, the model builds its answer from other sources in your category — and those sources are usually your competitors. So it is not hallucinating. It is saying what it can best support based on the public web. Which is not you.

In other words: “hallucination” is often not the model’s fault, but the imprint of your own information gaps. And this distinction matters, because it places the problem back where it can be solved: inside your own digital footprint.

How do language models actually “see” you?

To understand the problem, it is worth briefly clarifying how an LLM knows that you exist. Through three channels:

• Training data. During training, the model processed a massive amount of public text. If there is barely any public, well-written content about you, you barely exist in its “internal knowledge” either.
• Live search and browsing. Modern assistants search the web in real time before answering. Here, what matters is what bots and search engines can publicly find at the moment of the query.
• Structured data. Schema markup, clean headings, tables, and frequently asked questions help the machine not only read your content, but also interpret and extract it.

All three channels have one common denominator: publicly available, machine-readable information. What is not online or cannot be interpreted practically does not exist for the model. And this is where the two most common self-inflicted traps appear.

Problem 1: hidden pricing quietly removes you from the shortlist

The “let’s talk about pricing” tactic has its own business logic: the sales team wants to control the conversation and position the pricing range according to the buyer’s size. Ten years ago, this was a working strategy. In an LLM-driven buying process, however, it is an own goal.

When the buyer includes a budget in the question — “€4,000 monthly budget,” “small team, limited budget” — the model tries to recommend only options it can confirm will fit. If you have no public pricing, the model has nothing to confirm. It does not take the risk: it simply leaves you out and includes the competitor whose pricing is public, transparent, and categorizable.

The painful part is that this is not a loud rejection. Nobody tells you directly, “you were too expensive” or “we did not know how much you cost.” You are simply left out of the conversation. Quietly. Immeasurably — until someone checks how you appear in AI answers.

An important nuance: this does not mean you need to display every price in full detail. It is enough for the model to find a public, structured reference point — a starting package price, pricing tiers, “from” prices, or a transparent pricing logic — something it can hold onto. Complete opacity is the problem, not the existence of a sales process.

Problem 2: a closed feature = a non-existent feature

The second trap is even more subtle, because this is where you often bury your best capabilities. A feature that exists only behind a login, in a sales presentation, or in a downloadable PDF does not exist for the model.

Suppose your product is market-leading in real-time, two-way data synchronization. But there is no public, searchable, well-written page about this anywhere — it only appears on slide 14 of your sales deck. When someone asks AI, “which solution can do real-time two-way sync,” the model does not mention you. It mentions your competitor, because they wrote a public page about it — even if their solution is weaker than yours.

This is the ruthless irony of the SaaS world in the AI era: the better product loses to the better-documented product. The model does not evaluate code quality or user experience. It evaluates what it can find evidence for. If the evidence sits behind a login, it is not evidence to the machine.

And the effect compounds. The more competitors publicly document a capability, and the less you do, the more the model starts treating their version as the category’s default truth. After a while, you are not just omitted: the machine’s “norm” works against you, because your approach becomes the exception for which there is no public evidence.

Why does this hurt most with enterprise customers?

In the enterprise segment, the impact is more dramatic for two reasons. First, this is where deal values are highest, so every lost shortlist position directly represents serious revenue. A single unspoken recommendation can make a measurable difference years later.

Second, enterprise buyers are the most likely to use AI for pre-screening. They do not have the capacity to manually compare fifteen vendors, so they ask the assistant for a narrowed-down list based on criteria: compliance, integrations, security, pricing range. Exactly the criteria that hidden pricing and closed documentation prevent the model from checking for you.

In other words, you lose your most valuable customers in the least visible way. You do not receive an alert. It does not appear in a lost-deal report. Fewer qualified prospects simply reach you — and your marketing team thinks, “this month is weak.”

The buyer of the future does not ask AI whether your company is “good.” They ask which one is “the best” — and if you are not in the list, your answer was never born.

A five-minute self-test before you start anything

Before you begin any strategy or audit, it is worth seeing the problem with your own eyes. Open a language model — ChatGPT, Claude, Gemini, or Perplexity — and ask exactly the questions a real buyer would ask. Here are four that will give you an immediate picture:

• “What is the best [category] solution for a [size] company with a [budget] budget?” — Do you appear in the list at all? And if yes, in what position?
• “Compare [your brand] and [competitor].” — Does it describe your features accurately, or does it provide outdated, incomplete, or incorrect information about you?
• “How much does [your brand] cost?” — Can it say anything concrete, or only that the buyer should “contact the sales team”?
• “Which [category] tool can do [your key feature]?” — Does your name appear, or your competitor’s?

If you are consistently missing from the answers, placed behind competitors, or described inaccurately, then this is not a theoretical risk. Revenue leakage is already happening — you simply had not looked at where. This self-test is the raw diagnosis; the full, system-level picture and prioritized plan are what a thorough audit adds.

What can you do? Five steps toward machine visibility

The good news is that this is not black magic. Machine visibility can be optimized just like classic SEO used to be — only the target audience is different: now it is not a search engine, but a language model “reading” you. Here are the most important moves:

1. Bring pricing out of the dark. You do not need to publish a full price list, but provide a public, structured reference point: pricing ranges, a starting price, or a transparent pricing logic that the model can hold onto.
2. Document your features publicly. Every important capability should have its own standalone, searchable, well-written page. If it is behind a login, it does not exist for the machine. Bring it into the light.
3. Structure content for the machine. Clear headings, answer-first paragraphs, FAQ blocks, schema markup such as Article, FAQPage, Product, and Organization. Help the model not only read, but interpret.
4. Build comparison and evidence-based content. Case studies, specific numbers, integration lists, compliance proof. The model likes verifiable claims — give it plenty.
5. Measure how AIs talk about you. What you do not measure, you cannot improve. You need to know what ChatGPT, Claude, Gemini, and Perplexity currently answer to the key questions in your category — and where you stand in that picture.

This final point is the one most often skipped — and the most decisive. Most companies optimize blindly because they have never checked what AI actually sees. Yet this is exactly the starting point without which the other four steps are only guesswork.

This is where the AI Visibility Audit comes in

The AI Visibility Audit begins with what is most useful: we check how the four most important language models — ChatGPT, Claude, Gemini, and Perplexity — describe and rank you today, based on real buyer questions in your own category.

We map where you are invisible, where they describe you inaccurately, and exactly whom they recommend instead of you. We examine where your pricing leaks out of the buying journey, which of your features do not exist in the machine’s eyes, and which of your claims the model cannot verify from public sources.

Then you receive a concrete, prioritized action plan: what needs to be made machine-readable, where, and in what order — from pricing to feature documentation to schema markup — so that AI recommends you the next time a buyer asks. Not generic advice. Tailored to your category, your competitors, and your current visibility.

The goal is simple: do not let an opaque algorithm quietly decide who gets onto your buyer’s shortlist. You decide — by making yourself visible.

The audit result is not a forty-page PDF that ends up in a drawer. It is a transparent, prioritized list: which few changes will have the greatest impact on your visibility, and what can wait. This way, your team can start working on it not next quarter, but tomorrow.

Frequently Asked Questions

Do LLMs really recommend specific software?

Yes, and increasingly often. Buyers turn to language models not only for information, but also for comparisons and recommendations. The assistant typically names 3–5 solutions with explanations — and this becomes the buyer’s starting shortlist.

If I publish my pricing, won’t I lose sales control?

Not necessarily. You do not need a full, detailed price list, but a public, structured reference point — a pricing range, starting price, or transparent logic — that allows the model to categorize you. Personalized quoting can still remain with the sales team.

What is the difference between classic SEO and AI visibility?

Classic SEO optimizes for appearing higher in a search results list. AI visibility, often called GEO or AEO, optimizes for a language model mentioning you by name, accurately and positively, inside its answer. The two are connected, but they are not the same: AI does not provide a list; it makes a decision on your behalf.

How do I know whether this problem affects me?

The fastest test: ask ChatGPT, Claude, Gemini, and Perplexity yourself using the typical buyer questions in your category. If you do not appear, or they describe you inaccurately, you are seeing the problem live.

How long does it take to see results?

After public pricing and feature documentation are added, and structured data is implemented, models relying on live search can start mentioning the company differently relatively quickly. The effect built into training data is slower and longer-term — which is why it is worth starting as soon as possible.

Before your next enterprise customer asks ChatGPT, it is useful to know what it will answer.

With the AI Visibility Audit, you will discover where you stand today in the eyes of language models — and exactly what you need to do so they recommend you.