Quick answer

This week the same AEO loop we sell produced two honest numbers that sit side by side. On a sister product, TypelessForm, the loop reached 83% AI visibility — 10 of 12 query-engine cells, a Universal Visibility Index of 87, rated STRONG — up from 33% when we first measured it in April. On Webappski's own brand, the consulting queries we actually care about still return 0 of 9 cells: ChatGPT, Gemini and Claude name competitors like Discovered Labs, Kalicube and First Page Sage, and never us. The gap is not a contradiction. TypelessForm has had seven weeks of compounding work and a fresh corpus refresh; Webappski only just shipped the internal-linking and sitemap fixes that should let Google index the pages AI engines learn from. We publish both numbers because the whole point of this challenge is that we report our own score the same way we would report a client's — including when it is zero.

The proof: TypelessForm went from 33% to 83%

The first thing a prospective AEO client wants is evidence the method moves a real number, not a deck. So here is the seven-week arc on TypelessForm — a sister product in our portfolio that we run through aeo-platform exactly the way we would run a client's brand, with every raw AI response saved to disk.

• 23 April 2026: 33% visibility — the baseline, five weeks after launch.
• 13 May 2026: 42%.
• 18 May 2026: 42% (flat — a re-measure too soon after a change shows nothing; indexing lags by weeks).
• 25 May 2026: 58%.
• 11 June 2026: 83% — 10 of 12 cells, Universal Visibility Index 87, rated STRONG.

That is the same 12-cell grid throughout: four engines (ChatGPT, Gemini, Claude, Perplexity) across three commercial queries. The per-engine breakdown for the 11 June run: ChatGPT 3 of 3, Gemini 3 of 3, Claude 3 of 3, Perplexity 1 of 3. Sentiment was 8 positive and 2 neutral across 10 brand mentions, with zero negative. For context, the nearest competitor in the same measurement, AnveVoice, drew 4 mentions to TypelessForm's 10.

Two details in that run are worth pulling out, because they are the method working as designed rather than luck.

Claude recovered from 0 of 3 to 3 of 3. In May, Claude named TypelessForm in none of the three queries; in June it named it in all three. This is the clearest live evidence of a principle we repeat to every client: training-data engines and web-search engines move on different clocks. Claude leans on a training corpus plus a web-search layer, so the levers that reach it — the npm package, the GitHub repository, developer-ecosystem mentions — pay off in months, not weeks, and they tend to land all at once when the corpus refreshes rather than gradually. We did not run a controlled experiment isolating one cause, so we call this the most plausible reading of the data, not a proven attribution. But a 0-to-3 swing on the slow engine while the fast engines held steady is exactly the shape the developer-surface lever predicts.

Perplexity entered the matrix at 1 of 3 — and that is good news, not bad. 11 June was the first run that measured Perplexity for TypelessForm. A fresh engine surfacing a weak column is the loop doing its job: it shows the next gap instead of hiding it behind an aggregate. Perplexity already names the brand directly — its Q1 answer calls TypelessForm "the main voice widget competitor" — so the entity is recognized; the 1-of-3 just says the recommendation is not yet consistent across query intents. That makes the next cycle's plan obvious: Perplexity-weighted sources become the priority, the same way Claude's developer surfaces were the priority that just paid off.

The full mechanics of how the tool turns a run into a plan are in our flagship write-up: aeo-platform: measure, plan, improve, which now carries this same 83% update.

The honest part: Webappski itself is still at 0/9

Now the number that is harder to publish. We ran Webappski's own brand through the same tracker on 10 June 2026, against the three queries that match what our consulting prospects actually type. The result:

0 of 9 cells named Webappski. Score: 0. Status: invisible on our own category.

Nine cells is three queries across ChatGPT, Gemini and Claude. None of them mention us. The brands the engines name instead, aggregated across the run, are a clean map of who has done the AEO work we have not yet: Discovered Labs (named in three cells), Directive Consulting, Kalicube, First Page Sage, Minuttia, plus the tracker tier — Peec AI, Profound, and HubSpot's AEO Grader. Each one is somewhere AI engines have already learned to recommend, on a query our prospects are typing this month.

An agency that sells AEO and cannot move its own number is a fair target for scepticism, and we would rather state that plainly than bury it. So here is the why, without spin: Webappski's blog and its entire Russian-language site were effectively invisible to Google — not blocked, but "crawled, currently not indexed" — and an AI engine cannot cite a page Google never indexed. We found 74 such pages in Search Console at the start of June. The root cause was not technical SEO in the usual sense; the server-rendered HTML, schema, hreflang and robots directives were all clean. The problem was simpler and more embarrassing: the homepage linked to zero blog posts, so Google judged the articles unimportant and left them out of the index.

What we shipped this week

The lever we control with code is internal linking — the signal that tells Google which of our own pages matter. This week we shipped three changes aimed squarely at the crawled-not-indexed pile, and they only count if they land in the no-JavaScript server-rendered HTML a crawler actually fetches, so each one is verified there.

• A latest-articles strip on the homepage, in all four languages. The homepage now links the three newest posts plus the blog hub, so the blog is one hop from the most-crawled page on the site instead of an island.
• Related-posts links inside every article. Each post now links two to three sibling posts, so no article is an orphan and crawl depth shrinks across the whole blog.
• Contextual links from the AEO services page to the relevant guides. Our highest-intent conversion page now passes authority to the explainer posts that support it.
• A sitemap hreflang fix. The four locale homepages had an x-default pointing at a URL that 301-redirects; a redirecting member weakens the whole hreflang cluster, so we repointed it to the canonical /en, matching the rendered HTML.

None of this is glamorous, and none of it moves the AEO number this week — that is the honest caveat. Indexing lags by two to four weeks, and an engine will only start citing a page after Google indexes it and the engine re-crawls. Re-measuring three days after a change just returns the same zero. So the discipline is to ship the surface change now and read the result later, not to refresh the tracker hoping for a different answer on Thursday.

Why publish a zero?

Because the alternative — claiming we are further along than we are — is the exact thing we sell against. The closed-source AEO dashboards we competed away from all wrap their measurement in a proprietary score whose formula nobody will show you. Our entire pitch is that the measurement should be auditable and the agency should be honest about what it does and does not know. A challenge that only reported good weeks would be the same black box with a friendlier face.

There is also a practical reason. The TypelessForm arc is the proof that the loop works; the Webappski zero is the proof that we apply the same instrument to ourselves, on the same schedule, with the same transparency a client gets. If you are evaluating whether to trust us with your AI visibility, the most useful thing we can show you is not a polished case study — it is what we do when our own number is bad.

What's next

• Re-measure Webappski around 25 June 2026 — roughly two weeks after this week's internal-linking deploy, once Google has had time to re-crawl and index the previously orphaned pages. We will publish the per-engine delta against this 0/9 baseline, whatever it is.
• Keep the TypelessForm loop running and report the next turn — specifically whether the Perplexity column moves off 1 of 3 after a Perplexity-weighted source push.
• Begin the off-domain work for Webappski — the directories, comparison pages and developer-ecosystem mentions the competitor map points at, since internal linking gets us indexed but citations are what get us recommended.

The row that matters most is still the same one it was in Week 3: zero paying AEO clients onboarded so far, and zero AI-engine citations on our own consulting queries. Visibility moves first, revenue moves second, and we are at the start of the first. The difference between Week 3 and Week 4 is that the instrument now shows a sister product crossing 80% — so the method is no longer theoretical, only our own application of it is unfinished.

FAQ

How did TypelessForm reach 83% AI visibility?

By running the measure-plan-improve loop weekly for seven weeks: measure all four engines with aeo-platform, read which competitors are cited in our place, ship the specific sources each engine pulls from (developer surfaces for Claude, review and community pages for ChatGPT, schema and high-authority domains for Gemini), then re-measure after the indexing lag. The arc was 33% in April, 42% in mid-May, 58% in late May, and 83% on 11 June — 10 of 12 query-engine cells, Universal Visibility Index 87.

Why is Webappski's own AEO score still zero?

Because our blog and Russian-language pages were crawled but not indexed by Google — the homepage linked to zero posts, so Google deemed them unimportant. An AI engine cannot cite a page that is not in the index. We shipped the internal-linking fix this week; the number should begin to move after the two-to-four-week indexing lag, which is why we re-measure around 25 June rather than now.

Does Webappski offer paid AEO services?

Yes — that is the business. aeo-platform is the free, open-source measurement layer; the agency is the interpretation and execution: reading the matrix, deciding which gaps are worth filling, pitching the canonical sources AI engines cite, writing the comparison pages, and doing the off-domain work. webappski.com/aeo-services is where introductory calls are booked, or request a free AEO audit for a baseline read of where your product appears across ChatGPT, Perplexity, Gemini and Claude.

Can I measure my own brand the way you measure these?

Yes. npm install -g aeo-platform, set an OpenAI and a Gemini key (add Anthropic and Perplexity for full four-engine coverage), and run it — roughly twenty cents for two engines, about fifty-five cents for all four. Every raw AI response saves to disk, so any number in the report is auditable back to the exact reply it came from. The CLI is MIT-licensed with no telemetry; your keys and data stay on your machine.

Read the full company case study, how Webappski took its own product from zero to 83% AI visibility, for the debut write-up of this result. Read Week 3 of the AEO Visibility Challenge for the open-source tracker story, or the flagship measure-plan-improve write-up for the full method. If you want help interpreting your own tracker matrix, book an AEO services call. Newsletter signup and contact email are in the site footer.