Titles & AI Visibility: Titles, they matter

On 25 September every crawler started getting our page titles in the head of the page, where they belong, and most pages got titles that say what they are. In the first entry we gave the starting point: in the week before, Claude cited laup.ai in 15 of its 308 answers to questions that don't mention Laup by name.

We track it on our own Laup board: card #24 is the rewrite, #25 the same fix measured on Claude alone, #26 on ChatGPT, Copilot and Gemini. On launch day each card froze the two weeks before as its baseline, and every nightly run since adds to the after side.

Share of all answers that name Laup, the two weeks before the fix against the nine nights since:

And what we changed, in order:

Three engines up by about five points each, one flat. That is the result. The nightly run is at 03:00 UTC, so a change shows up the night after it goes live.

Since the fix, Claude cited laup.ai in 36 of 396 answers to questions without our name: 9.1%, up from 4.9% (the board above counts only answers that name a brand, which is why it says 9.3%). The best nights are the last three, and the questions it cites us on are the ones we built pages for, like “Hvor kan jeg kjøpe en GEO-audit for nettbutikken min i Norge?”: eight of nine nights.

It is not all the titles. On 30 September we unpaired our Norwegian pages from their English twins, which brought them back into Claude's searches, and the best nights come after that.

We expected little from Gemini, since its search barely shows our titles. Instead it went from at most one answer a night to two, three, three, five, four, seven and four, starting on 28 September, before the copy change and the unpairing. Google got our pages through Search Console from 25 to 28 September, five to eight a day. The timing fits the titles reaching Google, but we can't see inside Google's index the way we can see Brave's, so that part is our best explanation, not something we measured.

Copilot was the one we planned as the control. Bing already got our titles in the right place before the fix, so what reached it was the new wording, two rounds of it, and the 29 September copy change. For four nights Copilot edged up, from about one answer a night to two. Then it jumped to five and six on 30 September and 1 October, the first runs after the 29 September change, and has held at four a night since.

We expected Copilot flat. It moved as much as Claude. Either the new wording mattered more than we thought, or the copy change did, and this post cannot tell the two apart. That is one for the next experiment.

ChatGPT already names us more than the others, in about one answer in seven, and that has not moved: 14.4% before, 14.4% after. Its crawler gets the titles in the head now too, so either it has not come back for our pages yet, or titles were not what held it back. We will know more on 9 October.

A change only reaches an engine once its crawler has fetched the page. Here is what we submitted and when each crawler came. Bing reaches Copilot, Brave reaches Claude, and Google reaches Gemini.

Crawls are counted from our own request log by each crawler's published IP ranges, because the names they send get faked: one “Bravebot” wave turned out to be our own checks. Two things stand out. Google's 41-page day on 27 September came the night before Gemini started naming us more. And Brave fetched nothing after 1 October, so Claude's best nights, 3 and 4 October, came with no new fetches: whatever moved was already in Brave's index.

Several things changed in these nine nights, so this post cannot say whether it was the titles, the new wording, the front-page copy, or the hreflang unpairing and x-default that did it. The next entry is an experiment to separate them, after the two-week result on 9 October.

A note for anyone doing the same: we see results very quickly when we submit the changed pages to Bing, Brave and Google Search Console, instead of waiting for the crawlers to come on their own. More on that later.