Our page says 60+ languages. Crawlers read 0. AI apps often say 4

A number I never wrote

On 18 October 2026 I measured how AI apps describe M-Meeting, our meeting transcription and translation service, whose page lives at migurei.ai/m-meeting. I asked ChatGPT, Claude, Gemini and DeepSeek, in Japanese.

I expected to hear "60+". It is the big number near the top of the page.

Instead, on all four apps, many answers said M-Meeting supports "four languages: Japanese, English, Vietnamese and Korean". The sources they cited were our own press release from 24 September 2026 and sites that had republished it.

What a bot sees

The same day I fetched the page the way a crawler does: a plain request, no JavaScript. In the HTML, the language count reads 0, followed by +. In other words, 「対応言語 0+」.

The page's script starts at 0 and counts up to 60 in under a second. Anyone opening the page in a browser sees "60+". A crawler that does not run JavaScript sees only the starting value.

There is good reason to think most AI crawlers do not run JavaScript. A December 2024 study by Vercel and MERJ found that none of the major AI crawlers rendered it, apart from Gemini (which uses Googlebot's infrastructure) and AppleBot. Google's own JavaScript SEO guide still recommends server-side rendering or pre-rendering, because not every bot can run JavaScript.

The second surprise

When I went looking for the "right" number, I found that we ourselves were telling different stories. Visitors to the page see 60+. Crawlers reading the page see 0+. The press release says 4. The language list inside the M-Meeting app has 6: Japanese, Vietnamese, English, Korean, Chinese and Thai. Which one is correct is a call for the product team, not for this post.

But from an AI app's point of view, only one source stated a clear number it could actually read: the press release.

What to do with your site

1. Look at your page source, not the rendered page, and search for the numbers that matter: prices, customer counts, languages, founding year. If the HTML says 0 or leaves a gap, that is what crawlers see.
2. Put the real number in the HTML your server returns. If you want a count-up effect, let it animate towards a number that is already there; never keep the real figure only inside JavaScript.
3. Check your press releases and other pages that describe your product. When your own page gives crawlers no number, apps will take one from there.

How sure is this

This is a lab observation, without a control.

What is solid: the static HTML says 0+, read on 18 October. What is inference: that the apps quoted four languages from the press release because the page said 0+. They might cite the press release even if the page showed the right number. The Vercel and MERJ study is from 2024, and some vendors' crawlers may have started running JavaScript since; if so, they would read 60+, which still is not 4. This is one page; a second day of measurement showed the same "four languages" pattern. The full data is on the pw research pages.