The „Węże” lime kiln website had several dozen subpages and robust SEO, yet in June 2026, no AI model mentioned it in its responses to 18 questions about lime. After refining the existing content – without adding any new subpages – it appeared in every one of those responses by September. We’ll show you which aspects of SEO work carry over into AI responses, and what needs to be added.
Visibility in AI and SEO share a common foundation, but a good ranking on Google is not enough to ensure that AI chatbots mention the brand. In an Ahrefs study from early 2026, only 38% of the pages cited in AI Overviews appeared in the top 10 results for the same query. Our case study also illustrates this: Wapiennik „Węże” from Działoszyn had an extensive website with several dozen subpages, yet in June 2026 it did not appear in any of the AI’s responses to 18 questions about lime. After refining the existing content – without adding a single new subpage – by September it appeared in 18 out of 18 results, and the models cited its website for 94% phrases.
This text is part of the guide Brand visibility in AI: how it works and how to measure it. We have compiled the definitions of these terms and the differences between them in the text SEO, GEO and AEO: how do they differ?. Here, we’re focusing on the data: what aspects of our SEO work carry over into AI responses, and what needs to be added.
Both depend on whether a web crawler can access the page and index it. Both reward content that answers the question specifically, with figures and examples. Both use trust signals: links, mentions on other websites, and domain history.
Google states explicitly in AI function documentation, …that the same basic SEO principles apply to AI Overviews and AI mode as to regular search results. The page must be indexed and eligible to appear with a snippet. There are no additional technical requirements. A page that has an indexing issue with Google will almost certainly have the same issue in AI responses.
| SEO | Visibility in AI chats | |
|---|---|---|
| Result | the page’s position in the results list | reference about the brand or citation pages in the response |
| Starting point | keyword | a question phrased in natural language, broken down by the model into sub-queries |
| Stability | the position changes gradually | The composition and order of the list of brands change almost every time the programme is launched |
| What is assessed | Your website | everything the model reads about the brand, including on other people’s websites |
| Data | Google Search Console, tools for monitoring rankings | Search Console does not break down citations by AI features; Bing Webmaster Tools only shows them for Copilot and Bing – you’ll need to measure the rest yourself |
| Effect | clicking on the result | The user often receives a pre-written reply and doesn’t click |
TWO DIFFERENT ORDERS
The same foundation, a different result and a different level of stability.
A common foundation: indexing, specific content, trust signals.
aeloapp.ioThe biggest difference lies in consistency. On Google, a website ranked third today will most likely still be around third place tomorrow. In an AI chatbot, the same brand might appear first, fifth or not at all when the same question is asked again. In a study by SparkToro and Gumshoe.ai ChatGPT and Google’s AI produced the same list of brands in fewer than 1 in 100 attempts. See the text for details Why does AI give a different answer every time?.
It helps, but less and less. In an Ahrefs study from July 2025 76% of the pages cited in AI Overviews were in the top 10 for the same search query. In update from early 2026, based on 863,000 keywords and 4 million URLs, this percentage fell to 38%. The remaining citations were distributed almost equally between positions 11–100 (31.2%) and pages outside the top 100 (31%). Part of the decline is due to improved citation detection in the Ahrefs tool, so the two results are not fully comparable. BrightEdge, using a different method and different data, estimated the overlap with the top 10 at approximately 17% (a feature in Search Engine Journal).
SEO VERSUS VISIBILITY IN AI
Where in the Google search results were the webpages cited in the AI’s responses?.
Ahrefs study, 2026, 863,000 keywords and 4 million URLs.
aeloapp.ioIn ChatGPT, the mechanism is similar. In the AirOps survey of March 2026 32.9% of the cited pages appeared exclusively in the results of secondary queries that the model generated in the background, rather than in the results for the user’s query. A page may rank highly for the main search term but still lose out because the model was looking for something more specific: a price, a specification or a standard. The text describes how this mechanism works How AI chatbots choose the brands they recommend.
The „Węże” lime works is a family-run producer of building and fertiliser lime based in Działoszyn. Before the first audit, the website had a solid SEO foundation: separate pages for building lime and fertiliser lime, a page on standards, a blog and several dozen local subpages. Despite this, in June 2026, no search engine model mentioned the brand in response to 18 queries about its product category. Instead, the models recommended the two largest producers in the country.
The report revealed why. The models recognised the brand name of the flagship product, „Jurajskie Białe” lime, but as a type of lime rather than a specific company’s product. We describe this discrepancy between the product name and the brand in the entry brand aliases. In traditional SEO, this problem goes unnoticed, as a website may rank highly for a product name, yet the model will still fail to link it to the company.
The changes related to the existing content, not the number of pages. The page on standards now clearly states compliance with the PN-EN 459-1 standard. A table of specifications and a section on pricing were added to the building lime page, as the phrase combining the product name with the price and manufacturer had the lowest score in the first report. The weakest local subpages were tagged with the keyword „manufacturer” and given an indicative price. Not a single new subpage was created.
| Credits | 24 June 2026 | 27 July 2026 | 10 September 2026 |
|---|---|---|---|
| Search queries mentioning the brand | 0 out of 18 | 17 out of 18 | 18 out of 18 |
| Citation index | 6% | 50% | 94% |
| Brand mentions | 0 | 10 | 19 |
| Models who are familiar with the brand | 3 out of 4 | 3 out of 4 | 4 out of 4 |
| Models who endorse the brand | 0 | 0 | 1 |
Source: three Aelo reports across four models (ChatGPT, Claude, Gemini, Perplexity), all identical frozen query set, questions without a brand name (neutral scan). We publish this data with the client’s consent.
Three conclusions can be drawn from this case study for anyone who links SEO with visibility in AI:
Three reports from a single brand are not enough to distinguish the impact of changes to the website from fluctuations in product models and their updates. However, the trend is consistent across all metrics. You’ll find a full description in the two parts of the case study: From zero to 63% visibility in AI in five weeks i An AI model recommends a brand for the first time.
The model reads the text. A comparative article that describes the brand without a link still tells the model that the company exists, what it does and who its competitors are. In SEO, such a mention used to be of little value. It was precisely the lack of such independent signals that held up the recommendation in the lime manufacturer’s case for the longest time.
The 95% ChatGPT follow-up queries in the AirOps study had no search volume in keyword tools. Questions such as „does lime comply with the PN-EN 459-1 standard?” or „how much does a tonne of agricultural lime cost, including delivery?” will not appear in the keyword plan, yet these are precisely the kinds of questions the model asks in the background.
In SEO, it is enough for a webpage to rank highly for the product name. In its response, the AI model also needs to know whose product it is. The product name should appear alongside the company name in titles, descriptions and structured data, and the company description should be the same on the webpage, in Company Profile on Google and in catalogues.
For a long time, ChatGPT’s search mode used the Bing index (Lily Ray's analysis (it describes how this is changing), and Microsoft Copilot is also based on Bing. Many Polish websites do not even have a verified account on Bing Webmaster Tools nor the sitemap submitted there. From February 2026, the tool is set to AI Performance Report, which shows which Copilot subpages and AI responses in Bing are cited, and for which queries.
A robots.txt file that blocks an AI search engine bot prevents the site from being indexed, even if it ranks highly on Google. OpenAI describes its bots in documentation: OAI-SearchBot is responsible for searching within ChatGPT, whilst GPTBot handles the training data. In Vercel analysis None of the major AI bots execute JavaScript, so content loaded via a script may be invisible to them.
Not as two separate projects. Visibility in AI is an extension of well-managed SEO, incorporating elements that have previously been overlooked: measuring brand presence in the responses of several models, working on presence on third-party websites, specific facts in response to follow-up questions, indexing on Bing, and robots.txt settings for AI bots. The case of the lime manufacturer shows that often, rather than creating new subpages, all that is needed is a different perspective on the existing ones.
The first step is to check the initial state, as described in the text How to check whether AI recommends your brand. At Aelo, we measure this in the module Visibility: we run the same set of queries across four models (ChatGPT, Claude, Gemini, Perplexity) and show how a brand’s presence changes over time and relative to the competition. The results are fed into a knowledge graph Synapsis, and on this basis we draw up specific measures in Tickets. It’s up to you to decide which parts of this you’ll put into practice.
Agencies wishing to incorporate this area into their clients’ SEO services will find further details on the website Aelo for agencies, and owners of smaller businesses on the website Aelo for small businesses.
You can start tracking with a free account. Sign up
No. In an Ahrefs study from early 2026, 38% of the pages cited in AI Overviews were in the top 10 for the same query, whilst 31% came from outside the top 100 results. A high ranking increases the likelihood of being cited, but the model also compiles its response from related search results and from other websites mentioning your brand.
Not always. In the case of the lime manufacturer, not a single new subpage was created, yet the citation score rose from 6% to 94% in 11 weeks. The changes involved adding facts to the existing pages that the model can cite: standards, parameters and prices.
No. Indexing, accessibility for webcrawlers and substantive content are prerequisites for a website to be considered a source of AI answers at all. Google applies the same basic SEO principles to AI Overviews and AI mode as it does to regular search results. What changes is what you need to measure and what you need to include in your website optimisation work.
Either manually or using a tool that asks the same questions across several models at regular intervals. Since February 2026, Bing Webmaster Tools has been showing citations in Copilot and AI responses on Bing, but for ChatGPT, Gemini, Claude and Perplexity, no such report is available from the providers. The method for calculating the results is described in our methodology.
