Website citations in AI: the lime producer as a source for every question

This is the third part of a case study on a family-run manufacturer of construction and agricultural lime from the Łódź Voivodeship, which has been in operation since 1992. The first, From 0 to 17 mentions on AI in five weeks, describes the measurements taken in June and July, as well as the changes to the website that took place between them. The second, An AI model recommends a brand for the first time, shows the September reading and the first recommendation. Here, however, we are compiling the results of the scans carried out on 10 and 14 October.

How have the website’s citations in AI changed since June?

The citation index rose in each measurement, reaching its peak in October and maintaining that level in two subsequent readings: on 10 and 14 October.

Source: Aelo scans in four models: ChatGPT, Claude, Gemini and Perplexity. We are publishing this data with the client’s consent.

All the scans used the same frozen query set: 18 questions about building lime, agricultural lime, deliveries and a local manufacturer. We asked the questions without mentioning any brand names, that is to say, in neutral scan. Furthermore, the table only shows metrics that can be compared across all measurements. The method of calculation is described in methodology.

What does it mean when a page is cited in all search queries?

For each of the 18 questions, at least one model cited the manufacturer’s domain as the source of their answer. A ‘mention’ simply means that the model named the brand. A ‘citation’, on the other hand, means that they based their answer on content from the brand’s website and provided a link.

This is particularly important in this category, as the models almost always cite their sources. For example, in the October assessment, 71 out of 72 responses (99%) included a source. A customer who asks the AI about building lime or agricultural lime receives a reply containing links, and the manufacturer’s website is included among them for every query.

We calculate the citation rate against all phrases in the analysis, not just those for which the model cites anything. We explain in more detail in the text how the two metrics differ and how to check them manually How to check whether AI recommends your brand.

Which AI source is cited most often in relation to questions about calcium?

The manufacturer’s website, wapiennikweze.pl. The models cited it in 18 different phrases, whilst other domains in this category cited it in no more than 9. This is therefore twice as often as the next source, which includes agricultural websites, classifieds sites and other manufacturers’ websites.

Taką pozycję dały istniejące treści, uporządkowane w pierwszej części case'u: zapis zgodności z normą, parametry produktu i sygnał „producent” na podstronach lokalnych. Co więcej, w październikowym audycie strona dostała ocenę bardzo dobrego stanu technicznego i gotowości do cytowania. Tematy zakupowe są pokryte na tyle, że kolejne podstrony groziłyby już kanibalizacją fraz. Ten sam mechanizm widać w case'ie studia graficznego, opisanym w tekście AI chatbot recommendations after 5 months: the Click Made case: website citations there rose from 0% to 75%.

In this article, we explore why models choose certain sources whilst ignoring others How AI chatbots choose the brands they recommend.

In which areas does the brand perform best?

For building lime compliant with the PN-EN 459-1 standard and for our own product, „Jurajskie Białe” hydrated lime. These phrases have the highest visibility scores across the entire set:

The search result combines the brand’s presence in the responses from the four models with its position on the list (screenshot from 14 October 2026).

The two phrases containing the product name are the most important here. In June, the models were familiar with the name „Jurajskie Białe”, but treated it as a type of lime rather than a specific company’s product. The phrase linking the product to its price and manufacturer, meanwhile, achieved the poorest result. Today, both are at the top of the rankings, meaning the model knows whose product it is.

How does the brand compare with its competitors?

The „Węże” lime kiln appears in the results for all 18 search queries. The two competitors, whom the client identified themselves when launching the brand, appear in 33% and 17% search queries.

We are therefore looking at the same set of phrases, so the difference is clear: in every query in which any of these competitors appears, the manufacturer from Działoszyn also appears. ChatGPT and Claude also mention its name in every response.

We do not publish the names of competitors. The article explains how to compare yourself with the competition in AI-generated responses, and why this is a different metric to Google rankings Visibility in AI and SEO.

What next, to get AI to recommend the brand more often?

The brand’s recommendation in the AI rankings from September has held up: one in four models still single out the manufacturer as a good choice. For others to recommend it, confirmation from sources outside the brand’s own website is needed, as the website is already fulfilling its role.

This is because models are looking for independent sources that will say the same things about the brand as the brand says about itself. In this case, the plan involves:

• expert articles in the construction and agricultural media,
• inclusion in guides and round-ups from which models draw their lists of recommended manufacturers,
• customer reviews on independent websites, not just on the company’s Google Profile.

These tasks are available in the module Tickets, and we prepare the materials for them in Studio. The situation where a model is familiar with a brand but does not recommend it is described in more detail in the text Knows the brand, but doesn’t recommend it. We’ll show the results of our off-site activities in the next section of the case study.

How can you check a website’s citations in AI for your own brand?

It’s simplest with the Free account on Aelo: one brand, tracking across four models, and a full report including citations, mentions and competitor analysis. We analyse whether your competitors mention you, identify what’s preventing them from recommending you, draw up an action plan and prepare the materials needed to implement it. However, it’s up to you to decide what to make of it.

We explain what the free account includes in the text Free brand visibility measurement in AI: Free account. Agencies wishing to operate such a measurement for several customers, they can find details on the website Aelo for agencies, and independent businesses on the website Aelo for small businesses. The guide provides a broader overview of the topic Brand visibility in AI.

Try it for free

FAQ

Sources

• Aelo screenshots from 24 June, 27 July, 10 September, 10 and 14 October 2026: 18 queries across four models (ChatGPT, Claude, Gemini, Perplexity).
• Full brand visibility report dated 14 October 2026.
• Previous parts of the case: Part One and Part Two.