I'm confident it's not temp wikis. In editor_month, all of the temp users have the user_is_temporary flag, and our active editor query simply users WHERE NOT user_is_temporary in both of the appropriate places.

I also broke down new active editors by "main wiki" (the wiki where the new editor made the most edits). By far the biggest April–July changes (note: not the same as the year-over-year change) were at Wikidata and Commons, which haven't had temp accounts deployed yet.

Source code for this analysis.

thanks @nshahquinn-wmf , I checked YoY numbers using your code and the growth seems to be from Commons and wikidata as well.
In conjunction to the growth in Commons, I saw that net new content on Commons has increased, we usually attribute this to bot contributions but surprisingly this month we had a lot of new content added by non-bot users on Commons and Wikidata!

We continue to see these trends in August as well.
Wikidata seems to be pulling most of the weight in the new active editor growth in July and Aug 2025. here's why -
in July 2025, Wikidata accounted for more than half the growth in new active ed increases; and in Aug 2025 Wikidata increases were more than the overall new active editor increases.

raw numbers -
July 2025, had an increase of 2113 new active editors compared to july 2024, and wikidata saw an inc of 1204, while commons had an increase of 921 new active editors.
Aug 2025, had an inc of 849 new active editors compared to last yr; wikidata saw 116% inc ie 938 new active editors growth, while commons had an increase of 334 new active editors.

Further investigation:

1. new registered users on Wikidata has doubled in March 2026 compared to last year same time (wikistats). This aligns closely with a steep increase in desktop user pageviews starting 2026.

2. In addition to account registrations, new active editors and retained newcomers have also increased this month. However retained editors have reduced YoY.
   • definitions: data glossary
   • retained editors : logged-in users who made at least one edit in the previous month, the number who also edited during the current month)
   • Retained Newcomers: Out of the newly registered users who constructively activated (i.e. made at least one content-namespace edit during their first 24 hours after registration which was not reverted within 48 hours), the number who made an edit (in any namespace, whether reverted or not) during the days 2-15 following their registration.

These trends indicate a hypothesis that Wikidata could be affected by automated bot editing activity that isnt tagged as such because we don't have bot identifiers in contributors data currently.

Analysis:

• To verify this, I tried to connect newly registered wikidata users' first edit within 5 mins of registration with pageview data for the month of March. Looking at ip, user agent, x_is_browser score, accept_lang, referer etc.  their pattern seems pretty legit and human.
• we also see some newcomers (example) on Wikidata that are doing promotional editing for a political campaign, with a user agent on login that indicates AI agent usage (by searching for “Claude Code” or “OpenClaw” in the user agent header field in Logstash) . Most AI agents would disguise their user agent header but some of them don't.

Conclusion:

As of now we cannot definitively say if majority of the new registrations on wikidata are bot or not.
We should continue investigating this. perhaps expanding the search into all their revisions and identifying bot like patterns from the pageview data.

One hypothesis is that Wikidata is being increasingly edited by people for SEO/GEO/AEO purposes. There are industry articles that suggest people have their content on Wikidata, for example:
https://www.seostrategy.co.uk/wikidata-seo/ (published in Feb 2026, according to the source code)
https://cubitrek.com/blog/the-wikipedia-proxy-using-wikidata-ids-to-anchor-brand-truth/ (Dec 2025)
https://www.v9digital.com/insights/what-is-wikidata-and-why-does-it-matter-for-your-brand/ (updated Mar 2026)
https://www.beutlerink.com/ai-visibility-for-brands
https://www.luminodigital.com/ai-search-visibility

And Wikidata searches have increased according to Google Trends (https://trends.google.com/explore?q=wikidata&date=today%205-y&geo=US)

It corresponds with the anecdotal evidence of newcomers doing promotional editing.

Maya & I did some spot-checking of the accounts created in March; many of the ones we checked appear to be doing promotional editing:
https://www.wikidata.org/wiki/Special:Contributions/SpinexPickball
https://www.wikidata.org/wiki/Special:Contributions/Dinroampo
https://www.wikidata.org/wiki/Special:Contributions/Autonocion77
https://www.wikidata.org/wiki/Special:Contributions/Smithdentistry
https://www.wikidata.org/wiki/Special:Contributions/Aidentical
https://www.wikidata.org/wiki/Special:Contributions/DeMorais.art
https://www.wikidata.org/wiki/Special:Contributions/Dentalofficeofmoody
https://www.wikidata.org/wiki/Special:Contributions/Sandra-Recodex
https://www.wikidata.org/wiki/Special:Contributions/Identity-Team

After digging into the data further, the increase in Wikidata registrations appears to be related to a surge in promotional editing.

@JAnstee_WMF manually checked a random sample of 70 accounts that registered in March. Of those, 81% had edits that appeared to be either promotional edits or their edits were no longer available (due to edit/page deletions). "Promotional edits" were identified as such when they were about a single business or person (or a set of pages all related to a single business or person) that appeared obscure.

@Hghani queried the users who had registered in March; 4,342 had made at least 1 Wikidata edit. Of those, 1,975 (45%) of the users added a statement with an "official website".

337 of the March registered users have either been blocked or have a global lock on their accounts.

As I noted previously, there are a bunch of SEO/GEO/AEO industry articles that recommend making sure your information is on Wikidata - here's one that has a step-by-step guide: https://www.seostrategy.co.uk/wikidata-seo/ (it was published in February 2026, according to the source code). We're also seeing increases in Wikidata in Google Search trends – so I interpret this to mean that there's human interest (these aren't necessarily bots – and Maya's earlier investigation did not find clear evidence that the registration increase was driven by bots).

Thank you for the summary @kzimmerman. How confident are we in this conclusion, and is there anything still unclear?

@ARamirez_WMF The high proportion of accounts doing promotional editing is clear.

The driver for the promotional editing is not totally clear. Based on earlier analysis, the accounts did not seem to be coming from a single source. The spike does seem to correspond with recent SEO industry articles that suggest adding content to Wikidata. But it is possible that something else, or multiple factors, are driving the spike.

Thank you Kate. @Maria_Lechner_WMDE we'll want to discuss this at the quarterly meeting.

I believe we can close this after our latest discussion.

As discussed our next steps will be:

• Make it a topic in our upcoming conversations/research with UwER
• Continue to facilitate the RfC around notability policy reform
• Move the conversation about automatically identifying non-notable Items forward (T419106)
• Enable the new revert risk model on Wikidata (next step: T429049)
• Later: enable Automoderator (T365732)