Meta may be assembling its own web index so that Meta AI stops routing search queries through Google, according to a claim relayed by Search Engine Roundtable’s Barry Schwartz. The report rests on an August 6 post from developer Pieter Levels on X, not on any statement from Meta. That distinction matters for how much weight the claim deserves right now.
Levels wrote that Meta is allegedly building a search engine so its AI does not have to send web queries to Google, where those queries could feed Google’s own training data. He argued the company needs a web index it controls rather than one borrowed from a search partner, and he backed the post with screenshots he said showed a crawler hitting his own sites. Levels separately said an unnamed Meta employee contacted him directly and that he had permission to publish that exchange, though the message’s content is not described in Schwartz’s writeup.
None of this is confirmed. Schwartz’s account does not include a response from Meta, a named crawler user agent, a stated index size, or a launch date. Readers should treat the index claim as a report about one social media post, not as a Meta announcement.
The pattern Schwartz points to is not new. Facebook pursued its own search product for more than a decade, then leaned on a Bing arrangement to supply web results, an arrangement it walked away from roughly two years on, per Schwartz’s account. Building a proprietary AI stack now appears to be pulling Meta back toward the same ambition, this time to feed a chatbot rather than a social feed.
The mechanism Levels describes is worth naming directly. When an AI assistant answers a question by querying a rival’s search index, that rival sees the query stream: what users ask, how often, and in what phrasing. An owned index would remove that leak and turn Meta AI into a citation surface in its own right, rather than a reseller of somebody else’s ranked results. That is a materially different competitive position, one where a brand’s presence in Meta’s answers would depend on Meta’s own crawl and ranking choices, not on how it ranks in Google or Bing.
For site owners, the concrete change, if the claim holds, is another crawler to identify in server logs, another user agent to evaluate for access, and another robots.txt decision to make once the traffic pattern becomes routine. None of that can be acted on yet. Schwartz’s source material does not supply a crawler name to look for.
Search teams should also register what remains unknown. Schwartz’s report supplies no detail on how much of the web the alleged crawl covers, whether Meta AI already uses it for any live answers, or whether the activity shown in the screenshots is a test rather than a production system.
Search teams should watch server logs for unfamiliar crawler activity over the next quarter and treat any confirmed Meta AI citation surface as a distinct optimization target from Google’s, not an extension of it.
Barry Schwartz reported the claim for Search Engine Roundtable on August 10, 2026, citing an August 6 X post from developer Pieter Levels.
