Built with Croftweaver: one site, measured

The site

genai-wednesday.de is the home of GenAI Wednesday, a generative AI community in Munich: events, an archive of talks and a set of resource pages. The site launched on 27 March 2026 with these skills. The copy of the skills and tests in its repository is stamped with toolkit commit 41cc760, taken on 9 July 2026.

Speed and search basics

Performance

99

Accessibility

100

Best practices

100

SEO

100

Google's PageSpeed Insights, on a simulated phone, for the home page: one run on 4 October 2026. Scores run from 0 to 100 and vary a little from run to run. Re-run the test.

Clicks from Google since the launch

In the 28 days from 2 September 2026 to 29 September 2026, Google sent 50 clicks to the site, from 2,426 appearances in search results. The chart shows how that number moved. The table gives the clicks in separate 28-day blocks, counted back from that day. The launch on 27 March 2026 falls inside the first block.

Clicks rose from almost nothing to a best 28 days of 99 (16 June 2026 to 13 July 2026), and have fallen since. The cause of the fall is not established.

For its two main local searches, Search Console reports these average positions from 2 September 2026 to 29 September 2026: 8.9 for ai events munich and 8.0 for ai events münchen. An average position is the mean over every appearance, not a fixed rank. The links run the search for you; what you see depends on where you are.

Source: the site's own Google Search Console account, whole site, all countries. You cannot re-run these figures, so the daily numbers behind them are published: download the file (CSV). The latest block ends on the last day Search Console had data for when the file was made.

Do AI assistants name it?

On 8 October 2026 the AI check asked the assistants it ran two questions, up to three times each:

1. What are good AI meetups or events in Munich?
2. Is there a free generative AI community in Munich for founders and builders that meets regularly?

Google AI Mode and Google AI Overview were not run that day. The table shows their last results, where they have any.

With web search, 16 of 18 answers named the site that day. Without it, 0 of 24 did. The check counts mentions for two questions on one day, and answers change from run to run. It talks to each assistant through its API, which can answer differently from the app.

Why the second number is zero: an assistant recalls a name from memory mostly when it met that name often in the text it was trained on, and a name reaches that memory only when a new version of the model is trained. The site launched on 27 March 2026, which is recent for that. Web search does not rely on that memory, which is the likely reason the first number is higher. The check only counts who names the site. It does not test why, and it asks one model per assistant, not always the largest.

The counts behind the table are published: download the file (CSV, one row per day, assistant, mode and question). In the file, cited counts the answers that listed the site among their sources, whether or not the answer names it. For Perplexity those sources are the ones it returned, and the check does not tell which of them the answer quotes.

This is a different thing from the panel in the Builder Lab, where four assistants rate the site. A rating is an opinion; this check only counts whether the site is named.

What this page does not show

• More than one site. One example cannot show that a result repeats. More proof adds speed scores for three sites.
• A comparison. Nothing here shows what the site would have done without Croftweaver.
• Large numbers. This is a community site with tens of clicks a month, not a traffic story.

The Builder Lab keeps a fuller scorecard for the same site.