AEO and GEO case study: GeckoBoost measured by its own method, from zero

Why publish a case study about ourselves?

No client has agreed to a published case study yet, and we will not invent one. The only proof we can show honestly is our own: a brand that starts with no visibility in AI answers, a measurement method fixed in advance, and a public record of what changed. This page is that record. It is updated at every measurement point, including when nothing moves.

It is written for marketing managers and agency owners who want to see answer engine optimization (AEO) applied to a real brand before paying for it, and who prefer a dated number to a promise.

Where did GeckoBoost start on 10 November 2026?

We took the baseline before publishing a single optimized page. Two collections ran that evening: AI answers to 31 queries in English, French and Dutch, and Google results for 20 queries in two markets each, 40 result pages in all.

Two limits apply. Gemini was collected through a lightweight model, so its answers may differ from the Gemini app. Perplexity, Copilot and Claude were not measured that evening, so they have no baseline. If they are added to the tracking later, they will be reported as a separate series.

What did the baseline show about the market?

• No agency leads. On no query did all four engines name the same agency. The most named agency appeared in 9 answers across 6 queries.
• Tools are concentrated. Semrush was named in 25 answers, Profound in 22, Peec AI in 16 and Otterly.AI in 13. A new tool enters a crowded list.
• Social platforms are the top sources. YouTube was cited in 24 answers and Reddit in 20, mostly by Google AI Mode and AI Overviews. ChatGPT cited neither and preferred vendor pages and self-published “best of” lists.
• Each engine has its own map. Sources cited by two engines on the same query overlap little: an average Jaccard index of 0.08 over 154 engine pairs.
• Answers move fast. Between 31 October and 10 November 2026, four of the nine domains in the Google top 9 for “best AEO agency” changed.

What is GeckoBoost publishing, and in what order?

The plan has three waves. The first goes online before search engines are let in, so the site opens with pages that answer the questions buyers actually ask. Every page has a direct answer under its title, questions as headings, a visible FAQ, a named author and visible dates.

• Wave 1, before opening: the AEO and GEO services page, the AI visibility audit, the free AEO checker, the guide to answer engine optimization and the about page.
• Wave 2, weeks 3 to 6: the AI visibility tool, AEO vs GEO vs SEO, how to appear in ChatGPT, 10 actions to boost your AEO, what AEO and GEO cost and this case study.
• Wave 3, weeks 6 to 8 and later: the glossary, one guide per engine and product comparisons, each fact dated and sourced.

Each recommendation on these pages is tagged documented when an engine's official documentation supports it, with the link and the date, or heuristic when it comes from our own observations.

Which technical changes come with the opening?

• Open crawling and indexing in the same release (documented). To appear as a link in AI Overviews or AI Mode, a page must be “indexed and eligible to be shown in Google Search with a snippet” (Google Search Central, read 31 October 2026). OpenAI recommends allowing OAI-SearchBot (OpenAI crawlers, read 31 October 2026).
• Keep the full text in the HTML (documented). Bing lists “Hiding critical content behind client-side rendering” among the practices to avoid (Bing Webmaster Guidelines, read 31 October 2026). The site already renders every word on the server.
• Sitemap with exact lastmod dates, and IndexNow (documented). The same Bing guidelines name IndexNow, XML sitemaps and “accurate lastmod values” as discovery and freshness signals.
• Structured data that matches the visible text (documented). Google says it “isn't required for generative AI search” but still helps rich results (AI optimization guide, published 15 May 2026). We use it to describe pages, not to get cited.

And outside the site?

• Profiles under the final name on LinkedIn, YouTube, Reddit, G2, Sortlist and Clutch (heuristic). G2, Clutch and Sortlist all appear among the sources cited by the engines on 10 November 2026.
• Videos with transcripts (heuristic). YouTube is the most cited domain in our baseline.
• Real participation on Reddit, no planted mentions (documented). Google writes that “seeking inauthentic ‘mentions’ across the web isn't as helpful as it might seem” (AI optimization guide, published 15 May 2026).
• Entries in directories and lists that accept submissions (heuristic). ChatGPT favours self-published “best agencies” lists, and several of them take submissions.

How is progress measured?

The method is fixed before the first page goes live, so later readings compare like with like.

1. Six tracking projects, one per country

   United States, United Kingdom, France, Belgium, Luxembourg and the Netherlands. Belgium is measured in French, Dutch and English, Luxembourg in French and English.

2. 106 questions, asked every day

   54 distinct questions in English, French and Dutch, spread over three stages of a buying journey: discover, compare, decide. Each question gets one answer per engine per day.

3. The four baseline engines

   ChatGPT, Gemini, Google AI Overviews and Google AI Mode, the same four as on 10 November 2026. Perplexity, Copilot and Claude are not tracked for now.

4. Anchor questions kept word for word

   Five questions repeat the 31 October and 10 November collections exactly, such as “best AEO agency” and “AEO agency Belgium”, so the starting point stays comparable. The collection provider is not the same, and we say so with every comparison.

5. Competitors measured in the same answers

   Ten tools and the agencies named in each market, so share of voice is read against real names.

6. Google checked every week

   Each Monday, about 51 Google result pages: positions and the sources cited in AI Overviews, with the baseline settings. Google Search Console and Bing Webmaster Tools, including Bing's AI Performance report, start on the day the site opens.

What can a case study like this promise?

Nothing about rankings or citations. The engines say so themselves:

What we can do is measure, check eligibility and show change over time. Our working targets, such as being named by at least two engines on “AEO agency Belgium” within 90 days of opening, judge the plan. They are not commitments, and they are reported here whether met or missed.

Results so far: what has moved?

Each review adds a row with the same figures on the same sample, flat readings included. Market-level figures will also appear in the AI visibility barometer once the first market is published.

What can you take from this for your own brand?

• Measure before you publish (heuristic). Without a baseline, no later number means much.
• Measure several engines (heuristic). With an overlap of 0.08 between engines, a plan built for one engine leaves the others to competitors.
• Check eligibility first (documented). Crawler access, indexing and text in the HTML are the conditions the engines publish, with the sources quoted above.