Facebook Built 39 Million Answer Pages. You Cannot Copy the Recipe

Each page is a short summary built from posts other people wrote. Meta has published nothing about the directory. The pages keep ranking anyway.

The directory holds one page per question, and the questions run from card benefits to airport parking. Traffic grew faster than any content operation I have seen:

Month	Worldwide visits, Semrush estimate
May 2026	129,514
July 2026	1,032,527
August 2026	12,167,857
September 2026	36,388,769
October 2026	40,590,906

Saul Westbrook broke the story on 5 November 2026 and pulled the numbers. He also noted the keyword count behind those pages. It went from 4 million in July to 53 million in August, and Google holds about 6 million of the pages in its index.

The detail that matters most is how little editing went into it. Westbrook found 180 English pages on Padel Tennis, a sport with a small search footprint. Nobody checks these pages before they go live.

Full write up: https://www.peripheralcreative.com/news/facebook-answers-ai-generated-pages

Google's spam policy is public and blunt. Scaled content abuse covers a site that generates "many pages ... for the primary purpose of manipulating search rankings". It names generative AI tools as one way to break the rule, and it adds that the rule applies "no matter how it's created".

Read the policy: https://developers.google.com/search/docs/essentials/spam-policies

Westbrook points out that the pages came through the September spam update intact. I can't tell you why Google left them alone, and neither can he. What I can do is list what Meta had before it published a single page.

Meta owned the posts. Users have written on Facebook since well before 2010, and those posts are the source material for every summary. Meta owns a domain Google cannot demote without damaging its own results. Meta also owns the login, so a visitor who lands on one page becomes a session.

A small site has none of those three. That is the whole difference, and it has nothing to do with how clever the page template is.

One more piece Meta controls is the crawler that visits your site. Our guide on Facebookbot covers what it collects.

I opened a live page on 6 November 2026 rather than trusting a screenshot. The Amex travel page carries five headings of plain summary, and each claim leans on a named post.

Example: the Amex travel page

There is no byline, no publish date and no editor. A "More like this" block links to related pages inside the directory. Clickable sources sit under the summary.

That layout is the tell. The page is not written for a reader who arrived from a newsletter. It is written to survive one visit, collect a rating and pass the reader to another page.

We publish a lot of pages. In the last 90 days our Search Console shows 22 page and query rows containing "programmatic seo", spread across 10 pages. The best of them took 97 impressions at position 63, with zero clicks.

For "scaled content abuse" itself we have no rows at all. Nobody finds us on that term.

That is the honest picture of publishing at our size. Adding pages changes the count. It doesn't change the position, because the missing ingredient is the source material readers already wrote.

Our guide on programmatic SEO risks covers what goes wrong when a site builds pages without that material.

Almost every brand now has a summary about itself somewhere in results. Very few of those summaries are editable. This table separates the work that responds from the work that does not.

Surface	Who writes it	Can you change it	How
Your own site page	You	Yes	Edit the page
Your Facebook posts and comments	You	Yes	Post, correct, or add detail
A Facebook Answers page	A model, from its sources	No	Change the posts it quotes
An AI Overview citation	A model, from live results	No	Change the pages it quotes

The one route worth using is the third row. Your posts are quoted as sources, and sources get clicked. A plain sentence about your opening hours on your page is likelier to be quoted than a logo.

Run this search: site:facebook.com/fb-answers "your brand". Set the location to the United States, because the measured traffic sits there.

Open every page that mentions you and read the summary aloud. If a line is wrong, do not look for an edit button. There is none.

Then post the correction where the model reads. Put the right detail in a group post and in a comment under your own page. Say the fact in a full sentence, the way a person would write it.

Then compare the summary to your About page. If the two disagree, fix the About page first. That is the page you control.

One more check, and it costs nothing. Search your company name and see whether an answer page outranks you for your own name. That result tells you more than a rank tracker will.

Every figure here is a Semrush estimate, and Meta has said nothing. The 40 million visits may fall next month. Google may act on the directory after the public attention.

Reddit shows what a shift looks like when it lands. Its estimated search traffic fell to 1.8 billion visits a month from a 2.3 billion peak. Its share price dropped 29 percent at the end of October.

So treat this as evidence, not a forecast. Meta got a large slice of informational search by owning the posts and the domain. The lesson for a smaller site is the reverse of the tactic everyone is sharing.

Do not build 39 million pages. Check what the pages already say about you, and give the model better sentences to quote.

If a wrong line about your company is sitting in results today, book a call and we will find it: book a call