Guide: AI in business › Visibility in Google and in AI model answers
Optimising for ChatGPT means actions that increase the chance that ChatGPT will mention and cite your company when a customer asks about your sector: OpenAI crawlers’ access to your site, content that answers questions directly, and mentions in independent sources. A place in the answer cannot be bought. Nor can it be guaranteed. What you can do is honestly increase the chance and measure whether it works, and we will show where ChatGPT gets its sources from, what really helps and how much it costs.
In 60 seconds ChatGPT and your company
- What it’s about: about a brand’s presence in ChatGPT’s answers and, more broadly, about optimising for AI models, also known as LLM SEO.
- Who will find it useful: companies whose customers ask AI about suppliers and compare offers before sending an enquiry.
- Cost of measurement: a few hours a month manually, and answer-tracking tools from approx. USD 99 a month (September 2026).
- The first two steps: check that robots.txt isn’t blocking the OAI-SearchBot crawler, and record what ChatGPT says about your company today.
- What not to buy: ads in ChatGPT don’t affect its answers, and offers guaranteeing “first place” are selling a promise.
What is ChatGPT optimisation?
When a customer asks ChatGPT “which company in Poland can carry out an energy audit of a plant?”, they get a few names and a short justification. Optimising for ChatGPT is the work of making sure yours is among those names. And that the model tells the truth about your company. The scale is considerable. In February 2026, OpenAI reported that 900 million people use ChatGPT every week, and some of them ask exactly the questions they used to type into Google or put to a sales rep at a first meeting.
Several names for the same thing are doing the rounds in the industry: LLM positioning, AI model optimisation, LLM SEO. LLM (English: large language model, i.e. a large language model) is the technology that powers ChatGPT, Gemini, Claude and other assistants. Anyone asking “what is LLM SEO optimization” will get a simple answer from us: it is SEO extended to cover everything that shapes the answers of language models. The background, i.e. how SEO differs from GEO and AEO, is set out in a separate article on the three abbreviations, and ChatGPT’s place in the bigger picture is shown in our guide to AI in marketing and PR.
Where does ChatGPT get its information about companies?
ChatGPT has two sources of knowledge. The first is training: what the model learned from texts collected before its training cut-off date. The second is live search. According to OpenAI’s help documentation, ChatGPT itself decides when a question needs up-to-date information, rewrites it into one or more queries and also sends them to third-party search providers, among which it names Microsoft.
To show pages in search answers, OpenAI uses its own crawler, OAI-SearchBot. The company states plainly that pages which block it in their robots.txt file will not appear in ChatGPT’s search answers. There is also a second crawler, GPTBot, which collects data for model training; blocking it simply means “don’t learn from my content” and does not exclude the site from search.
Which sources does ChatGPT draw on? According to a Muck Rack study from May 2026 (more than 25 million links analysed), ChatGPT cites sources in 96% of answers, but only five on average. 84% of all citations in the models studied came from earned media, i.e. publications the brand did not pay for. So you are competing for one of a handful of places, and the strongest card in this contest is not your website but what others have written about your company: trade journalists, authors of round-ups, organisers of competitions and rankings.
| OpenAI crawler | What it is for | What blocking it in robots.txt means |
|---|---|---|
| OAI-SearchBot | showing pages in ChatGPT search answers | the site will not appear in ChatGPT search answers |
| GPTBot | collecting data for model training | content won’t be used for training; doesn’t exclude it from search |
| ChatGPT-User | visiting a page at a user’s request | robots.txt rules may not apply, because the action is initiated by a human |
Who ChatGPT optimisation pays off for
Not every company needs to compete for ChatGPT’s answers with the same commitment. Those whose customers ask advisory and comparative questions stand to gain the most:
- B2B companies with a long buying process. The customer gathers information for weeks and increasingly starts with a conversation with AI.
- Specialist services (consulting, technology, industry, logistics), where the customer asks “who in Poland does…”.
- Brands entering a new market that the model doesn’t know anything about yet.
- Companies that have changed their name or scope of business, about which ChatGPT still repeats old information.
- To a lesser extent local everyday services, because there the decision is made in Google Maps and reviews, not in a conversation with a model.
How to rank in ChatGPT step by step
This plan works in companies without a dedicated SEO team. The order matters. Each step builds on the previous one.
- STEP 01Record your baseline
List 20–40 customer questions and ask them in ChatGPT in a logged-out window. Note whether the brand is mentioned, with what source, in what tone, and whom the model names instead of you. Choose a few control questions that you won’t touch.
- STEP 02Open the door to crawlers
Check your robots.txt. OAI-SearchBot must have access, and while you’re at it, make sure your most important pages are indexed in the search engines ChatGPT uses.
- STEP 03Rewrite the pages you want cited
The first sentence answers the question directly, headings read like customers’ questions, figures have a source and a date, and the author is named. A model is more likely to cite a passage that can be lifted out without context, so a paragraph starting with “As we all know, the market is changing rapidly” will lose to one that says in its first words what you do, for whom and how quickly.
- STEP 04Earn mentions in independent sources
Expert comments and articles in trade media, presence in rankings and directories. This is a job for PR, not SEO, and according to Muck Rack data it matters most for citations.
- STEP 05Make your company data consistent
Your name, range of services, address and year of founding must read the same on your website, in directories and company profiles, and in your press office. Conflicting data is the most common cause of errors in answers, because a model that sees three different versions in three places picks one of them or blends them into something that matches none.
- STEP 06Measure again after 60–90 days
The same questions, the same method, a comparison with the control questions. You can check traffic from ChatGPT in your analytics via the utm_source=chatgpt.com tag, which OpenAI adds automatically.
What works and what is a myth
Plenty of promises have grown up around ChatGPT optimisation. This overview is based solely on what can be verified in sources, namely OpenAI’s documentation, Google’s guide and citation studies, not on what sales pitches promise.
| Claim | State of knowledge (September 2026) |
|---|---|
| “You can buy a place in the answer” | No. OpenAI says that ads don’t influence ChatGPT’s answers; they are labelled as sponsored and kept separate from the answers. |
| “An llms.txt file will get you into ChatGPT” | No public data shows that OpenAI uses this file; OpenAI’s documentation refers to robots.txt and OAI-SearchBot. |
| “Good SEO is enough” | Partly. SEO gives access to search, but ChatGPT mainly cites sources independent of the brand. |
| “Content with quotations and figures is chosen more often” | This is supported by a Princeton and IIT Delhi study (KDD 2024): such changes increased visibility in answers by up to 40% under experimental conditions. |
| “Once you have a result, it stays” | No. Answers change from day to day; what counts is the trend in repeated measurement. |
How much does ChatGPT optimisation cost, and what drives the budget?
The budget depends on the number of questions in the measurement, the amount of content to improve and the number of publications to secure. Below are public tool prices (as of September 2026) and the factors that shape the cost. We don’t compare agency service prices. The scopes vary too much.
| Item | Cost or factor | What it affects |
|---|---|---|
| Manual measurement in a spreadsheet | PLN 0 for tools, a few hours of work a month | enough to start with 20–40 questions |
| Semrush AI Visibility Toolkit | from USD 99 a month per domain, 25 questions, billed annually | daily tracking of ChatGPT, Gemini, Google AI and Perplexity |
| Ahrefs Brand Radar AI | from USD 199 a month as an add-on | mentions of your brand and competitors in AI answers |
| Rewriting pages | number of pages, access to in-house experts | whether passages on the page are fit to be cited |
| Media coverage | number of expert topics and trade media outlets | presence in the sources ChatGPT cites most often |
When to start and when to wait
Start when your sales reps hear customers say “ChatGPT recommended…” and it isn’t your company. A good moment is also the quarter before a launch, a new market entry or a sales season, because publications secured now will start working in answers with a delay.
Wait if your site has indexing problems or your services are described in a single sentence. In that case, sort out your SEO first, because without it OAI-SearchBot and the partner search engines simply won’t find content worth citing. We explain how to fix the basics and what stays the same in classic SEO in the article on how AI is changing SEO.
Gemini, Claude, Perplexity and Google: LLM optimisation more broadly
ChatGPT is the biggest, but not the only, model your customers talk to. Models differ in how they cite: according to Muck Rack, Gemini cites sources in 82% of answers, eight on average, while Claude does so in 55%, but when it does cite, it gives 13 sources on average. In Google, meanwhile, there are AI Overviews and AI Mode, which use a search engine index, so classic SEO carries the most weight there.
The good news? The foundation is the same. A site that crawlers can access, content that answers questions directly and mentions in independent media help with all models at once, which is why you should plan LLM optimisation as a single programme, not as separate projects for each tool with a separate budget, a separate report and a separate agency. We show how media coverage translates into AI answers in the article PR AI. We describe working with journalists itself in the article on media relations.
How to choose an agency for ChatGPT optimisation
The term LLM SEO agency (English: LLM SEO agency) is now used to describe an SEO firm, a PR agency or a specialist GEO team alike. Ask each of them three questions. How will you measure the baseline? How will you tell the effect from chance? Who will secure media mentions, and how? If the answer to the last question is “we’ll write sponsored articles”, remember that according to Muck Rack, paid content accounts for 0.3% of citations.
Avoid offers that guarantee “first place in ChatGPT” and packages built solely on an llms.txt file. You will find fifteen questions worth asking before signing a contract in the article AI agency or GEO agency: how to choose? If you’d rather measure everything yourself first, use our guide to auditing your visibility yourself.
Commplace® works on AI visibility at four levels and combines SEO with media relations. You will find the details on the page AI search optimisation. Join us for a free initial consultation.
Where companies waste time and budget in the battle for ChatGPT
Blocking all AI crawlers in robots.txt “just in case” excludes the site from ChatGPT’s search answers. Don’t want to train models? Block GPTBot. Leave OAI-SearchBot alone.
Nobody controls ChatGPT’s answers. According to OpenAI, ads don’t either. When you pay for a guarantee, you are paying for a promise that cannot be verified.
ChatGPT cites five sources on average, and most citations come from earned media. Even the best company website without independent mentions often loses out to a competitor that the trade press has written about.
According to Muck Rack, paid content accounts for 0.3% of citations in AI answers. Budget spent on it to “get into ChatGPT” usually doesn’t pay off.
Answers depend on how the question is phrased, the user’s history and the day. A single screenshot can send your budget in the wrong direction, and another one taken a week later with the same question may show something completely different, so base decisions on a series of measurements, not on a single image.
The ChatGPT questions we most often receive by email
What does ChatGPT optimisation involve?
On increasing the chance that ChatGPT will mention and cite the company when answering customers’ questions. It has three elements: OAI-SearchBot’s access to the site, content that answers questions directly, and mentions in independent sources, mainly the media. The effect is measured by repeated measurement against a fixed list of questions.
Can you pay for a place in ChatGPT’s answer?
No. OpenAI states that ads in ChatGPT do not influence answers: they run in a separate system, are labelled as sponsored and are kept separate from answers. Being in the answer itself depends on the sources the model considers credible.
How can I check whether ChatGPT recommends my company?
Ask ChatGPT, in a logged-out window, 20–40 questions your customers ask, and note whether the brand is mentioned, with what source and in what tone. After a month, repeat the measurement with the same questions. You will recognise traffic from ChatGPT to your site in analytics by the tag utm_source=chatgpt.com.
What is LLM SEO?
LLM SEO is the work of increasing a company’s visibility in the answers of large language models such as ChatGPT, Gemini or Claude. It combines classic SEO with work on mentions in independent sources. It is simply another name for GEO.
How long does ChatGPT optimisation take?
Judge the first changes in answers 60–90 days after improving your pages and securing publications. Answers can be fickle. The trend counts, not a single result.
Will blocking GPTBot remove me from ChatGPT?
No. GPTBot collects data for model training, while OAI-SearchBot is responsible for showing pages in ChatGPT’s search answers. By blocking only GPTBot, you exclude your content from training but can still appear in ChatGPT’s search results.
Does llms.txt help with ChatGPT optimisation?
There is no public, verifiable evidence that OpenAI uses the llms.txt file when answering questions. OpenAI’s documentation talks about robots.txt and the OAI-SearchBot crawler. You can have the file as a cheap extra, but not as the basis of your work; more about it in the article llms.txt: what it is and whether it works.
Sources
- SOURCETechCrunch: ChatGPT reaches 900M weekly active users · 27 February 2026
- SOURCEOpenAI: Searching the web with ChatGPT (help centre) · as of 24 September 2026
- SOURCEOpenAI: Overview of OpenAI crawlers (OAI-SearchBot, GPTBot, ChatGPT-User) · as of 24 September 2026
- SOURCEOpenAI: Publishers and Developers FAQ · as of 24 September 2026
- SOURCEOpenAI: Ads in ChatGPT (help centre) · as of 24 September 2026
- SOURCEMuck Rack: “What Is AI Reading?” — May 2026 edition · 7 May 2026
- SOURCEAggarwal et al., “GEO: Generative Engine Optimization”, arXiv / KDD 2024 · 16 November 2023
- SOURCESemrush: AI Visibility Toolkit pricing · as of 24 September 2026
- SOURCEAhrefs: pricing and the Brand Radar AI add-on · as of 24 September 2026
Read next
We measure your brand’s presence in ChatGPT and other models on your customers’ questions, improve your pages, earn media mentions and check the effect after 90 days.
Sebastian Kopiej, CEO of Commplace®. In public relations since 1996. Written with the help of AI tools and editorially verified by the author. Data current as of 24 September 2026.