Webappski's own agency brand is cited in just 2 of 39 AI-answer cells, and we publish that raw starting line. The same AEO method took our sister product to 83% AI visibility first. Both wins here are native-language — German on Claude, Polish-intent on Gemini — while ChatGPT names us nowhere. Publishing the low number is the point.

This is the second chapter of a public experiment. The first chapter — how we took TypelessForm from zero to 83% AI visibility — proved the method works on a product we control. This chapter turns the same playbook on the harder, more honest target: our own consulting brand. We ran our open-source tracker against thirteen native-language buyer queries across ChatGPT, Gemini, and Claude — thirty-nine answer cells in total — and we are reporting the result exactly as the engines returned it, including the cells where we do not appear at all.

Most agency content shows you the wins and quietly drops the losses. We do the opposite, because our entire offer is measured AI visibility, and a number you cannot reproduce is just marketing. The raw responses from this run are saved to disk and can be re-checked. If we cannot prove our own starting line, we have no business selling yours.

What Does "Turning the Playbook on Ourselves" Mean?

It means measuring our own agency's AI visibility before optimizing it, then publishing that first measurement unedited. Answer Engine Optimization (AEO) is the practice of structuring content, technical infrastructure, and online presence so AI answer engines — ChatGPT, Perplexity, Gemini, Claude — cite and recommend a company when buyers ask relevant questions. Unlike SEO, which optimizes for Google's ranking of links, AEO optimizes for the synthesis step, where a large language model decides which brands to name in a single generated answer.

A "raw starting line" is the baseline measurement taken before any optimization — the honest day-one number. For a brand, it is the count of AI-answer cells (one buyer query times one engine) in which the brand is actually cited. Ours, on 2026-06-14, was 2 of 39. The method that produced it is the same one we ran on TypelessForm: define the real buyer queries, ask each engine, record verbatim which brands it names, and store the raw output for audit. The only thing different here is the subject — us.

How Did a Sister Product Reach 83%?

Briefly, because the full story is its own article: our sister product TypelessForm — a voice-input form widget on a brand-new domain, no clients and no reviews — went from a zero baseline to a unified visibility index of 83, cited in 10 of 12 answer cells in our 2026-06-11 run, through structured content, schema, entity building, and per-engine optimization. We measured it rather than asserted it. The per-engine breakdown, the honest Perplexity caveat, and the dip-and-recovery are all in the zero-to-83 case study; this article is about what happens when we point that exact apparatus at ourselves.

What Is Our Own Raw Starting Line?

Our own agency brand was cited in 2 of 39 AI-answer cells on 2026-06-14, an overall score of 5%. We ran thirteen buyer queries — in English, Polish, German, and Russian — across ChatGPT (gpt-5-search), Gemini, and Claude, for thirty-nine cells. Two came back with a citation. Both were positive, and both happened to be in the native language of a buyer we actually want: one German, one Polish-intent. Here is the top of the raw summary file the tracker wrote to disk, unedited:

{
  "date": "2026-06-14",
  "brand": "Webappski",
  "domain": "webappski.com",
  "score": 5,
  "mentions": 2,
  "total": 39,
  "errors": 0,
  "generatedBy": "aeo-platform@1.3.2"
}

Do not take our word for the count — read the grid. We hosted the full tracker report, exactly as the open-source tool generated it, with only our internal file paths and per-run cost telemetry stripped: view our live AEO report → It shows all thirteen queries across ChatGPT, Gemini and Claude — every one of the thirty-nine cells, the two that cite us and the thirty-seven that do not, the competitor names the engines actually returned, and our own-domain signal audit. The article tells you 2 of 39; the report shows you the raw grid behind it.

The two citations are best read as the engine's own answer ordering in a single run, not as a stable ranking. In our 2026-06-14 run, Gemini's answer to the English query "answer engine optimization agency Poland for B2B SaaS" placed us first in its list — the raw response literally opens that section with "#### 1. Webappski (Gdynia, Poland)" and describes us as "AI-native, developer-first... an AI-search visibility studio founded by product engineers" who "open-sourced aeo-platform." In the same run, Claude's answer to the German query "beste Answer Engine Optimization Agentur für Sichtbarkeit in ChatGPT" placed us sixth, written verbatim as "### 6. Webappski (DACH-Region)." These are prose orderings from one engine on one day; between runs they move, so we report them as "in this run, the engine placed us Nth," never as a position we hold.

Now the gap. ChatGPT cited us in 0 of its 13 cells, across every language. That is the honest hole: gpt-5-search leans heavily on sources like Reddit and YouTube for this category, and we are not present there yet. So our 2 of 39 is not evenly spread — it is two native-language wins on Gemini and Claude, and a clean zero on ChatGPT. A standing scoreboard would hide that shape; the raw cells show it plainly.

If you read our Week-4 challenge note, you saw a different number — 0 of 9 — and the two are the same brand, not a contradiction. Four days earlier, on 10 June 2026, we measured Webappski on a deliberately narrow grid: three English consulting queries across ChatGPT, Gemini and Claude — nine cells — and scored a clean 0 of 9. That run is reported in Week 4 of the AEO Visibility Challenge. This run widens the grid to thirteen queries across four languages — thirty-nine cells — and catches our first two citations, both in regional, non-English cells. The English head did not move: ChatGPT is still 0, and the English-on-Gemini-and-Claude cells are largely empty too. What changed is not that the English number improved; it is that we measured the right surface — the regional, native-language one our buyers actually use — and the denominator grew with it. It is the same honest move the zero-to-83 case study made when it reframed a 100% headline down to 83% on a fuller grid: a bigger, more truthful denominator is worth more than a flattering small one. So read 0 of 9 and 2 of 39 as one timeline on two grids — the narrow English snapshot on 10 June, then the wider multilingual read on 14 June — not as two scores fighting each other.

• Cited (2 cells, both positive): Gemini on "answer engine optimization agency Poland for B2B SaaS" and Claude on "beste Answer Engine Optimization Agentur für Sichtbarkeit in ChatGPT." Our domain appeared in both answers' cited sources.
• Not cited (37 cells): every ChatGPT cell (0 of 13), and the remaining Gemini and Claude cells across English, Polish, German, and Russian queries.
• Why it still counts: the two wins are in buyers' native languages, on the exact regional intent we are built for — a thinner, more reachable answer space than the saturated US-English head.

Why Are We Visible in German and Polish but Not in US English?

Because the native-language answer spaces are less crowded, and because we measured them on real buyer intent rather than on the English acronym. The single most useful thing this run taught us is a translation trap: in Polish, "AEO" already means Authorized Economic Operator, a customs-and-trade term. A literal query for "AEO agency Poland" would measure customs brokers, not AI-visibility studios. So we built the Polish and German queries on how buyers actually phrase the intent — "widoczność w AI" (visibility in AI), "optymalizacja pod ChatGPT" (optimization for ChatGPT), "pozycjonowanie w AI" — not on a three-letter acronym that means something else in that language.

That intent-first approach is exactly why Gemini and Claude could find us in the right context: the engines answered the Polish and German questions in the buyer's language and pulled from a thinner pool of regional sources, where a well-structured local page earns a citation faster than in the crowded English space. The US-English head, by contrast, is dense with established US agencies, and ChatGPT's Reddit-and-YouTube-heavy sourcing for that head does not yet include us. The result is a map, not a verdict: we are reachable where our buyers actually are, and invisible where we are not yet trying to compete.

Who Gets Cited Today in Poland and DACH?

In this run, the recurring agency names were Delante and Agencja Whites in Poland and SearchGPT Agentur in Germany. Broader players like Omnius and Profound also surfaced. We extract competitor names directly from the engines' answers, so this is who the models actually name for these queries — not a list we picked. The table below pairs each with where it is based and what it leans on, so a buyer comparing options can see the field as the engines describe it.

The source pools matter as much as the names. For German queries, searchgptagentur.de appeared as a cited source in six answers; for Polish queries, regional sites like agencjawroclawska.pl and coderspilot.com recurred. These are the surfaces the engines trust in each market — and the concrete places a regional AEO plan has to earn a presence to move a 2 into a larger number.

What Is the Regional Plan to Move 2 of 39?

Appear where we are already nearly visible, re-measure after the indexing lag, and publish the raw number every time. The plan is deliberately narrow because the opportunity is regional, not global. We are reachable today on German (Claude) and Polish-intent (Gemini) queries, so the first move is to strengthen exactly those native-language surfaces — structured German and Polish pages, the regional source pools the engines already cite, and clean entity definitions an engine can trust when corroborating signals are sparse.

1. Deepen the native-language front. Build and structure German and Polish content where the engines already surface us, targeting the intent phrasings buyers actually use ("widoczność w AI," "KI-Sichtbarkeit"), not the English acronym.
2. Earn presence in the cited source pools. The engines pull DACH answers from sites like searchgptagentur.de and Polish answers from regional sources — a regional plan has to show up in those pools, not just on our own domain.
3. Re-measure on a respectful cadence. AI indexing lags by roughly two to four weeks, so we re-run the same thirteen queries after that window — measuring is a readout, not a lever — and report whether the German and Polish cells grew.
4. Treat ChatGPT's 0 of 13 as its own problem. Its Reddit-and-YouTube-heavy sourcing needs a different, community-led approach, tracked separately from the on-site work.

This is the work we do for clients — done-for-you AI visibility, optimized for each engine separately and measured with the same open-source tracker we just pointed at ourselves. If you want to see your own raw starting line before committing to anything, that is where we start too. See how our AEO services work.

Frequently Asked Questions

What is a "raw AEO starting line"?

It is the baseline measurement of a brand's AI visibility taken before any optimization — the honest day-one number. We express it as the count of AI-answer cells (one buyer query times one engine) in which the brand is actually cited. Our own agency's raw starting line on 2026-06-14 was 2 of 39 cells.

Why publish a low number like 2 of 39 instead of hiding it?

Because our offer is measured AI visibility, and a result you cannot reproduce is just marketing. Publishing the raw starting line — including the cells where we do not appear — is the proof that our numbers are real and audited rather than asserted. An AEO agency that hides its own baseline cannot credibly promise to measure yours.

Why is the brand visible in German and Polish but not in US English?

AI engines answer in the language of the question and cite sources in that language, and the German and Polish answer spaces are less crowded than US English. We measured on real buyer intent in those languages, so Gemini and Claude found us in the right regional context. The US-English head is denser and, for ChatGPT, sourced heavily from Reddit and YouTube, where we are not yet present.

What does "AEO" mean in Polish, and why did it change our queries?

In Polish, "AEO" is widely used to mean Authorized Economic Operator, a customs-and-trade designation. A literal query for "AEO agency Poland" would therefore measure customs brokers rather than AI-visibility studios. We built the Polish and German queries on intent phrasings buyers actually use — "widoczność w AI," "optymalizacja pod ChatGPT" — which is what let the engines surface us in the correct context.

How was 2 of 39 measured?

We ran thirteen native-language buyer queries across three AI engines — ChatGPT (gpt-5-search), Gemini, and Claude — which is thirty-nine answer cells, using our open-source tracker, aeo-platform. For each cell, the tracker records whether our brand is named in the answer and saves the verbatim response to disk. Two cells came back with a positive citation; the raw files for all thirty-nine are retained for audit.

Can I see the raw data, and can I measure my own brand the same way?

Yes. The measurement layer is our open-source tracker, which records which engines cite a brand for a given query and stores the raw responses, so the method is fully reproducible. You can run the same category questions in Polish, German, or English and note whether you are cited — or have us turn that into a clear baseline. Our AEO services start from exactly that measured starting line.

Conclusion: An Honest Starting Line Is the Product

We proved our AEO method on a product we control, taking TypelessForm from zero to 83% AI visibility. Turning the same method on our own agency brand returned a far humbler 2 of 39 — and that number, published unedited, is the most honest sales argument we have. It shows the method is real, the measurement is auditable, and the wins are exactly where a Polish or DACH buyer would look: German and Polish answers, on Gemini and Claude.

If you sell to buyers who research in Polish or German, the question is not whether AI answers already shape their shortlist — they do — but whether you know your own number. Start where we started: with a measured baseline that tells you which engines cite you today and which do not. See how Webappski's AEO services work, and we will show you your raw starting line before anything else.

This article was published in June 2026 and reflects a single tracker run dated 2026-06-14. AI answer engines are volatile — citations and answer orderings change between runs and over time, so the figures here describe that run, not a standing position. The raw responses are retained for audit. If you spot anything that has dated, contact us at info@webappski.com.