Meta built a directory of AI-generated pages, let Google index about six million of them, and the search industry has settled on a response within a day: name it scaled content abuse, screenshot the index count, and wait for Google to fix it. The naming is correct. The waiting is the part worth arguing with.
Facebook Answers sits at facebook.com/fb-answers/. Every page is an AI-generated summary of one topic, assembled from public Facebook posts and pages, and the directory is open to indexing by default. Saul Westbrook of Peripheral Creative flagged it first, writing in June 2026 that Meta had quietly indexed millions of AI-generated Facebook pages, each using an AI-generated summary to explain a specific topic.
Glenn Gabe put numbers on it on November 5, calling the directory another example of massive scaled content abuse. He reported that Facebook began populating it in July 2026, that it now holds more than six million indexed URLs, and that it ranks for millions of queries across organic results, AI Overviews and AI Mode, with a second post showing citations in Gemini as well. Barry Schwartz covered the whole thing at Search Engine Roundtable, noted rumors that Instagram is doing something similar, and added that he suspects this content will not stay indexed for long now that it is public.
Schwartz is probably right. Our argument is not that the forecast is wrong. It is that acting on it is the only move available to a publisher, and almost nobody is treating the directory as a thing to respond to rather than a thing to watch.
What Scaled Content Abuse Looks Like At Six Million Pages
Facebook Answers is a directory where each page is an AI-generated summary of a single topic, built from public Facebook posts, and Google has indexed roughly six million of them across a four month period. The mechanics are worth being precise about, because the scale is the least interesting part.
Each page summarizes content that is already public and already indexable on its own. That is the structural problem. A Facebook post that ranks is a primary source. An AI summary of several Facebook posts, published on the same domain, is a second indexable layer sitting on top of content that was already there, and it competes with the originals as well as with everyone else.
Google's own position is not ambiguous about the pattern. Its spam policy defines scaled content abuse as many pages generated for the primary purpose of manipulating search rankings rather than helping users, and its guidance on generative AI content says that using such tools to produce many pages without adding value for users may violate that policy. Google is also careful to say it judges purpose rather than page count, which is exactly the clause that makes this case harder than the screenshots suggest.
| WHAT GOOGLE'S POLICY OR GUIDANCE SAYS | WHAT THE FACEBOOK ANSWERS DIRECTORY DOES | OUR READ |
|---|---|---|
| Scaled content abuse is many pages generated primarily to manipulate rankings rather than help users | Roughly six million topic pages generated from post summaries, ranking for millions of queries | Fits on volume and generation method. Intent is the only genuinely arguable column |
| Using generative AI to make many pages without adding value for users may violate the policy | Each page summarizes public posts that are already indexed and already rank on their own | Added value is thin by construction when the sources are already in the index |
| Google judges by purpose, not page count, so a large site of useful pages is fine | Purpose is unstated, and Meta has said nothing publicly about the directory | The clause that makes the outcome uncertain rather than a formality |
| Platforms should decide whether a user-created page is meant for public discovery, and noindex it when it is not | The directory is open to indexing by default | Google published that platform guidance on October 26, ten days before this surfaced |
| Algorithmic spam systems and manual actions run on timelines Google does not pre-announce | Still indexed and still ranking as of November 5 | Nobody outside Google can date the correction, which is the practical problem |
Row four is the uncomfortable one. Google widened its user generated spam guidance on October 26 to cover exactly this situation, telling platforms to decide whether a generated page exists for public discovery and to apply a noindex rule when it does not. We went through that change in detail when Google expanded its user generated spam rules to platforms and app builders. Ten days later the largest social platform in the world had six million generated pages in the index.
Why The Deindexing Forecast Is Not A Plan
Google has published no statement about Facebook Answers, which means the enforcement timeline the commentary assumes is entirely unannounced, and no publisher can schedule a recovery against an action nobody has confirmed will happen or dated. That is not a quibble about certainty. It is a planning problem.
“A forecast about somebody else's enforcement calendar is not a strategy. It is a reason to do nothing, dressed as analysis.”
Consider how this plays out for a site that lost a handful of question queries to the directory in September. The deindexing-forecast position says: do nothing, the correction is coming, your rankings return. If that is right, the site recovers on an unknown date and has spent the interval earning nothing. If it is wrong, or if the correction is partial, the site has spent a quarter losing a query it never tried to defend.
We have watched the shape of this before. When Google rolled the September 2026 spam update out in phases, the practitioner instinct was to wait for the dust to settle before changing anything, and the sites that recovered fastest were the ones that had kept publishing through it. We wrote that up at the time in our read on the phased rollout of the September 2026 spam update. Enforcement events reward the people who were already working, not the people who paused to watch.
There is an honest counterargument and it deserves stating properly. If Schwartz is right and the directory is gone within weeks, then doing nothing is the cheapest correct answer, and any work you did was wasted effort on a problem that solved itself. That is a real position. It fails on asymmetry: the work we are recommending is a page you should have published regardless, and it keeps earning after the directory goes, so the downside of acting is near zero while the downside of waiting is a quarter of lost compounding.
Why Scaled Content Abuse Reaches The AI Answer Layer Differently
Removing a page from the Google index and removing it from the retrieval sets that ground AI Overviews, AI Mode and Gemini are two different operations, and the second has no published decay curve that anyone can plan against. This is the part of the story the index-count screenshots miss entirely.
Gabe's second post is the important one. He reported the directory showing up not just in organic results but in AI Overviews, AI Mode and Gemini, with citations captured in screenshots. Those are different retrieval paths with different refresh behaviour, and a URL leaving the organic index does not mechanically remove the claims it contributed from the answers already being generated.
We are deliberately not putting a number on how long that lag runs, because nobody has published one worth citing. What is documented is that the citation layer behaves as its own system rather than as a view onto the ten blue links, which is the case we made when we looked at where AI citations actually come from and how earned media feeds them. A model grounded on a page it can no longer fetch is a known failure mode, not a hypothetical.
The practical consequence is that a publisher tracking only organic positions will read the correction as complete before it is. That split, organic index running on one clock and the AI answer surfaces running on another, is why generative engine optimization services exist as a separate line of work rather than a bullet inside an SEO retainer. If you are measuring one surface you are measuring the easy half.
How we would triage a question query that a platform directory has taken. Our recommended order of operations, not measured data
The spam report sits at the bottom on purpose. It costs five minutes and it is worth filing. It is simply not a plan, and treating it as one is how a quarter goes by.
The Query Surface That Got Taken Was Available
Six million thin pages did not outrank established publishers on hard commercial terms, they occupied question-shaped long tail queries where no credible answer page existed, and that distinction is the whole lesson. A summary of Facebook posts does not beat a good page. It beats no page.
This is why the volume number is a distraction. The directory did not win on authority or on quality. It won on coverage, by generating a page for a question at a scale no publisher staffs for, and collecting the queries that were sitting unclaimed. Every one of those queries was a question a real audience was asking and nobody had answered properly on a page built to be found.
None of this is new advice. It is the same argument for writing pages that answer things properly, which Google itself keeps restating in its guidance on what counts as helpful main content, something we unpacked when Google tightened its definition of people-first main content. The directory is just an unusually clear demonstration of what happens to the queries you leave open.
What Should You Do If Facebook Answers Outranks You Now?
The correct response to a platform directory taking one of your question queries is to publish a better answer page this month, not to file a report and monitor, because the page keeps earning whether or not the directory survives December. Start with the queries, not with the outrage.
Concretely, over the next two weeks: export the question queries where a facebook.com/fb-answers/ URL appears in your results, discard the ones that were never going to convert, and pick the three that map to something you actually sell. Write one page per query, answer-first, with the answer complete in the opening sentence of every section. Then check those three queries in AI Overviews and AI Mode separately from your rank tracker, and keep checking after the directory disappears, because that is when you find out which surface was ever really yours.
For most teams this is a content problem with a measurement tail, which is where our content marketing and generative engine optimization work meet in practice. The page gets written once. The question of whether it is being cited gets asked every month, on surfaces that do not share a refresh schedule.
The broader read on this is less about Meta than it looks. A platform with unlimited generation capacity pointed it at an open query surface and took millions of queries in four months. Whether or not Google removes the result, the method is now demonstrated, cheap and public, and the next directory will not be announced by a practitioner on X with screenshots. It will just be there.
The queries that are safe are the ones with a real page behind them. That was true before July and it will be true after the correction, and it is the only part of this story you control.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.