Skip to content

From 2.8% to 28.7% in AI answers, in 96 days

A Swiss agency multiplied its visibility in AI answers by more than ten. The method: it found out where the models read and made sure it was listed exactly there. What we took from the Ranketta × Biggie webinar.

Dan Cristian Alexandrescu4 min read
Event
Become the Product AI Recommends: Agentic Commerce Playbook
Organizer
Ranketta AI
Speakers
Vojtěch Oravec (founder and CEO, Ranketta), Jan Synek (Biggie)
When and how
30 September 2026, 17:00–18:00 (EEST) · online webinar

We joined the webinar Ranketta AI ran on what actually gets recommended by ChatGPT, Google AI Overviews and Perplexity, not just tracked in them. One concrete case, with numbers.

What stayed with us after an hour: AI recommendations don’t come from everywhere. In the case shown, roughly 70% of citations came from a small group of sources. Whoever gets onto them has a good chance of getting into the answer.

In short
03
  • Biggie Switzerland went from 2.8% to 28.7% visibility in AI answers to “marketing agencies in Switzerland”, between 18 March and 22 June 2026, and now ranks 4th.
  • The method: they mapped the URLs the models cite, saw that roughly 70% of citations come from a small group of sources, and made sure the brand appears exactly on them.
  • Every platform has its own list. In Ranketta’s data, Reddit is in the top 3 on all five platforms, and YouTube in the top 5 on all five.
— The result

More than ten times as visible, in 96 days

18 March 2026 · start

2.8%

visibility in AI answers

22 June 2026 · after 96 days

28.7%

more than ten times the starting point

Current position

#4

in the ranking shown in the talk

The question measured: “marketing agencies in Switzerland”. Several AI models were queried, among them Google AI Overviews and Perplexity. Figures presented by Biggie, Ranketta webinar, 30 September 2026.
— The method

Three moves, in this order

04
  1. 01

    They measured what AI cites

    They mapped the exact URLs the models cite for digital-marketing questions in Switzerland. Not what AI says about them, but where it gets the information.

  2. 02

    They found the few sources that matter

    Roughly 70% of citations came from a small group: directories (sortlist.ch, leadingswissagencies.ch) and reference or niche pages (wikipedia.org, flin.agency, hey-uni.de).

  3. 03

    They got onto them

    They made sure Biggie appears on those sources. In the webinar they showed their Sortlist profile and their Leading Swiss Agencies profile, the second most-cited source.

  4. Note

    What the presentation doesn't show

    In the webinar screenshot, Biggie's Sortlist profile had 0 customer reviews. From what we saw, the result is given as a total, not per source: we don't know how much of the climb each one brought.

— Where each model reads

Five platforms, five source lists

Top five cited sources per platform, out of 25 million sources analysed by Ranketta
RankChatGPTGeminiAI ModePerplexityAI Overview
1WikipediaRedditGoogleYouTubeYouTube
2RedditYouTubeYouTubeRedditGoogle
3TechRadarWikipediaRedditGoogleReddit
4ForbesMediumFacebookWikipediaFacebook
5YouTubeGoogleInstagramLinkedInInstagram
  • Reddit: top 3 on all five platforms
  • YouTube: top 5 on all five

Wikipedia shows up on three platforms out of five, Google on four. We haven't seen the list broken down by market, so we read it as a general tendency, not as a rule for Romania.

— And here at home

We measured the same thing on the Romanian market, with our own data. In the book-market study, Reddit is the second most-cited source out of 422 domains (34 of 180 answers). In the electronics and IT one, it is the second most-cited source that isn’t a shop. Reddit ranks high here too, but the rest of the top is local: bookshops, retailers and Romanian articles. That’s why the list has to be built on your market and your field.

— Our takeaway

It's no longer only about how many links you have

The AI recommendation doesn’t start on your website. It starts from what the model reads about you elsewhere. That’s why in AEO the question has moved from “how many links do I have?” to “which cited sources am I on, how accurately and how verifiably?”

Models read context: who you are, what you do, whether what you say matches the rest of the web. A clean, honest presence on a few trusted sources weighs more than dozens of appearances with no weight. That’s our reading, not the speakers’.

— To do on Monday morning

Four steps, in order

  • Write the questions that bring customers

    10–20 questions, exactly as your customer would ask an AI. Whole questions, not keywords.

  • See which URLs the AI cites for them

    Not what it says about you, but where it gets it. Sources repeat: in Biggie's case, roughly 70% of citations came from a small group.

  • Pick 3–5 sources where you can genuinely be present

    Directories, reference pages, communities. Complete profile, correct data, real reviews. No inflated profiles: what you write there has to match the rest of the web.

  • Repeat on each model, separately

    ChatGPT, Google AI Overviews and Perplexity don't cite the same things. The webinar's advice: ChatGPT first, then Google AI Overviews. Start from a baseline measurement, as they did on 18 March.

How we read the number. One case, one question, a small market (Switzerland) and 96 days. The figures are the ones Biggie presented; we haven’t verified them independently. Visibility in answers doesn’t automatically mean leads or sales. And Ranketta’s list is a general tendency, not a rule for Romania.
— FAQ

Questions we got after the event

03
  • How much did Biggie climb in AI answers, and how fast?

    According to Biggie's presentation, from 2.8% to 28.7% visibility in AI answers to “marketing agencies in Switzerland”, between 18 March and 22 June 2026 (96 days). On 22 June the brand ranked 4th. The figures are Biggie's; we haven't verified them independently.

  • Where do AI models cite most often?

    It differs by platform. In Ranketta's data (25 million sources analysed), ChatGPT is led by Wikipedia, Gemini by Reddit, Perplexity and Google AI Overview by YouTube, and AI Mode by Google. Reddit is in the top 3 on all five, YouTube in the top 5 on all five.

  • Does this mean links no longer matter?

    From what we saw, the webinar didn't compare links against anything else. It showed that presence on the few sources AI cites moved the number. Our conclusion is more measured: the number of links is no longer a sufficient yardstick. Where you are cited, how accurately and how verifiably matter too.

— Surse

Where the numbers come from

03

The numbers on this page come from the presentations at the Ranketta × Biggie webinar on 30 September 2026, except for the two Websem studies cited in the text.

  1. 01
    Biggie2026

    Results: Visibility of Biggie Switzerland

    Webinar slide. 2.8% (18.03.2026) → 28.7% (22.06.2026), current position #4. Field: marketing agencies, Switzerland.

  2. 02
    Biggie2026

    Source discovery: where models look

    Webinar slide. Roughly 70% of citations from a small group of sources; directories and reference pages; advice: ChatGPT first, then Google AI Overviews, segment by model.

  3. 03
    Ranketta2026

    Top 5 most cited sources per platform

    Webinar slide. 25 million sources analysed; ChatGPT, Gemini, AI Mode, Perplexity, AI Overview.

— AEO

Want to see where AI reads in your industry?

We run the webinar's exercise on your company: the questions that bring customers, the sources the models cite and where you're missing. We start with a 30-minute conversation.