How to get AI to recommend your business: what works in 2026

Poniżej opisujemy, jak sprawić, żeby AI polecało firmę, na danych z badań i naszego case'u. Nie ma tu skrótu: polecenia w odpowiedzi nie da się kupić, a efekty przychodzą po tygodniach. Wapiennik „Węże” z Działoszyna w czerwcu 2026 roku nie pojawiał się w żadnej odpowiedzi na 18 pytań o swoją kategorię, a po 11 tygodniach był w każdej z nich i dostał pierwszą rekomendację.

This text is part of the guide Brand visibility in AI: how it works and how to measure it.

Where should you start to get AI to recommend your company?

Start by checking where you are. There are three possible states, and each requires a different course of action:

• The model does not recognise the company: he does not even mention it when asked directly; here, he starts from the basics, as described in the text ChatGPT doesn’t know my company,
• The model is familiar with the company, but does not recommend it: recognises the brand when asked about it, but does not name it when asked about the category; this is the most common scenario, as described in the text Knows the brand, but doesn’t recommend it,
• The model recommends the company, but rarely, or not in all chats: the aim here is to build on what is already working.

The difference between these states is explained by the term assessment and recommendation. You can check the status manually, as described in the text How to check whether AI recommends your brand, or in a free measurement on the Free account.

Why does the AI know about the company but not recommend it?

Because recognition and recommendation are two different things. To recognise a brand, it is enough for a user to have read about it somewhere. To recommend it, they must consider it a better fit for the query than others, and this requires confirmation from several independent sources.

The figures show this. In a study by SparkToro and Gumshoe.ai The lists of recommended brands differed almost every time, but the leading brands in the headphones category appeared in 55-77% responses. The model selects from a pool of brands it associates with the category. The whole process of making a recommendation boils down to getting into that pool. The text describes how the model builds it How AI chatbots choose the brands they recommend.

How do you prepare a webpage so that AI can quote from it?

The model looks for facts it can cite. When it responds using a search, it breaks the question down into several sub-queries: about price, specifications, reviews and availability. In the AirOps survey of March 2026 95% of such queries had no search volume in keyword tools, and ChatGPT cited only 15% of the pages it had retrieved.

The website should therefore include:

• price or at least the price range and what it depends on,
• specifications, standards and certificates, expressed in numbers rather than adjectives,
• for whom is a service or product for some, but not for others,
• region delivery terms and conditions,
• the answer at the start of each section, in a sentence that can be quoted without the rest of the text.

W research by academics from Princeton and IIT Delhi, which introduced the concept of GEO; adding statistics, quotations and sources to the text increased its visibility in generative responses by as much as 40%. We refer to this area as AEO.

This was the case with Wapiennik. The website had several dozen subpages, but they lacked specific details. The changes were made to the existing content: a clear statement of compliance with the PN-EN 459-1 standard, a table of specifications and a section on pricing. Not a single new subpage was created.

How do you link a product name to a brand?

This is a problem that isn’t apparent in traditional SEO. In the first survey, the respondents were familiar with the name of Wapiennik’s flagship product, „Jurajskie Białe” lime, but treated it as a type of lime rather than a specific company’s product. The brand had a recognisable product, which the model failed to associate with its name. This phenomenon is described by the term brand aliases.

The solution is simple, but it requires consistency. The product name should appear alongside the company name in page titles, descriptions, and structured data and in every external description of the brand. A single sentence such as „Jurajskie Białe is lime produced by Wapiennik Węże in Działoszyn” is more effective than several paragraphs on quality.

Where else does a company need to have a presence apart from its own website?

Where the model seeks confirmation. Most commonly, these are:

• comparisons and rankings in your category,
• business directories and Company Profile on Google, with the same details as on the website,
• customer review websites,
• the media and industry websites.

The simplest way to know where to start: check which sources the models cite when they recommend your competitors. This is a list of places where your competitors are present, but you are not. In the Wapiennik case, it was precisely the lack of such independent signals that held up the recommendation for the longest time, even though the website’s citations were growing rapidly.

At Aelo, we organise these off-site activities within Outreach: we advise where to set up a profile and on which platforms the brand should have a presence. Changes to the website itself are implemented in Tickets.

What about technology: AI bots, Bing and JavaScript?

Even the best content won’t be cited if a web crawler doesn’t read it. Three things to check:

• robots.txt cannot block AI search engine bots; OpenAI describes its documentation, by separating the ChatGPT search bot (OAI-SearchBot) from the training data collection bot (GPTBot),
• Bing index: ChatGPT relied on it for a long time, and Microsoft Copilot still uses it today, so the page should be verified in Bing Webmaster Tools,
• HTML content: in Vercel analysis None of the major AI bots executed JavaScript.

Google writes in AI function documentation, …that there are no additional technical requirements for AI Overviews and AI mode beyond standard SEO. However, he explains why SEO alone is not enough, Visibility in AI and SEO.

How long does it take before AI starts recommending a company?

There is no single timeframe, but the Wapiennik case illustrates a typical pattern. The first report, dated 24 June 2026: the brand scored 0 out of 18 responses. The second, dated 27 July: 17 out of 18. The third, dated 10 September: 18 out of 18, with the website cited in 94% phrases and the first recommendation in one of the four models. The citations came quickly, but the recommendation only arrived after 11 weeks. A similar pattern, albeit in the services sector, was seen with the graphic design studio described in the text AI chatbot recommendations after 5 months: the Click Made case: from zero in May to a recommendation from three AI engines at the end of October.

Changes to the sources that the model searches on the web take effect more quickly than changes to the knowledge acquired during training, which is only updated with new model versions. Read more in the text Why do changes take time? and in the second part of the case study: An AI model recommends a brand for the first time.

What shouldn’t you do?

When working on how to get AI to recommend a company, the following mistakes are the most common:

• Don’t count on being able to buy a recommendation. Advertisements in ChatGPT are labelled, displayed separately and, according to OpenAI, do not affect the content of the responses. See the text for details Advertisements on ChatGPT in Poland: what this means for brands.
• Don’t judge the results based on a single screenshot. ChatGPT and Google’s AI produced the same list of brands in fewer than 1 in 100 attempts. Read more in the article Why does AI give a different answer every time?.
• Don’t churn out content without any facts. The problem with Wapiennik was not the number of subpages, but the lack of specific details on those that already existed.
• Do not change the set of phrases being measured whilst you are working. Without frozen query set You won’t be able to compare the results before and after the changes.

How does Aelo help with this work?

To know how to get AI to recommend your company, you first need to identify what’s preventing it from doing so. We measure whether ChatGPT, Claude, Gemini and Perplexity mention your brand, identify what is preventing a recommendation, draw up an action plan and prepare the materials needed to implement it. Measurements in the module Visibility They are processed daily, and the results are fed into the knowledge graph Synapsis, and these form the basis for tasks in Tickets and materials in Studio. It’s up to you to decide which parts of this you’ll put into practice. The text describes the whole process An AI-powered brand visibility platform: how Aelo works.

Agencies carrying out this work for clients can find out more on the website Aelo for agencies, and local businesses on the website Aelo for small businesses.

Try it for free

FAQ

Sources

• SparkToro and Gumshoe.ai, a study on the consistency of brand recommendations in AI (January 2026): https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/
• AirOps, report on searches, follow-up queries and citations in ChatGPT (March 2026): https://www.airops.com/report/influence-of-retrieval-fanout-and-google-serps-in-chatgpt
• Aggarwal et al., GEO: Generative Engine Optimisation (KDD 2024): https://arxiv.org/abs/2311.09735
• OpenAI, robot documentation: https://developers.openai.com/api/docs/bots
• Vercel, analysis of AI robot behaviour (December 2024): https://vercel.com/blog/the-rise-of-the-ai-crawler
• Google Search Central, AI features and your website: https://developers.google.com/search/docs/appearance/ai-features
• Ahrefs, update to the citation study in AI Overviews (2026): https://ahrefs.com/blog/ai-overview-citations-top-10
• Aelo reports for Wapiennik „Węże” dated 24 June, 27 July and 10 September 2026 (published with the client’s consent)