Signal & Science
Signals and moves2026-09-21 · 9 min

Meta starts charging Pages to post links, and Google rewires the holiday funnel

Facebook link posts, Google's UCP cart transfer, Cloudflare's new robots.txt writer, and five more shifts that change what an operator should do this week.

Facebook Pages now capped on link posts unless you pay

Meta One for Business went live on September 15. The Advanced plan lists links in organic posts and reels as a feature, and page managers this week are reporting a two-link-post-per-month cap on non-paying professional accounts. Publisher Pages are exempt. Meta has not confirmed a full launch, but the wave is expanding beyond the December 2025 test.

This hits any small business, agency-managed brand, or fractional-run Page that has been quietly using Facebook to push traffic to a site, a booking page, or a landing page. If you post links more than twice a month, you are now choosing between a subscription, moving the URL into the first comment, or dropping Facebook as a link channel.

The move this week, under an hour: pull the last 90 days of Page posts, count link posts per month, and decide. If you are under two a month you do nothing. If you are over, either commit to comment-based URLs and rewrite your post template today, or price the Advanced tier against what those link clicks actually convert to. Do not subscribe on reflex.

Social Media Today: Facebook Pages get charged for link posts

Google Merchant Center adds cart transfer back to your site

Google is rolling out cart transfer inside the Merchant Center hub for the Universal Commerce Protocol. Shoppers who build a cart on Google surfaces, including AI Mode and Gemini, can be handed off to the merchant's own checkout. The Cart API is only available on UCP version 2026-04-08 and later. Rollout is U.S. first, with Australia and Canada scheduled for early next year.

This hits U.S. ecommerce operators who are already in or considering Google's on-Google checkout. Without cart transfer, you lose the sale entirely when a shopper abandons Google's native checkout flow. Cart transfer allows you to recover those shoppers to your own site.

The move this week: open Merchant Center, check your UCP version, and if you qualify submit the interest form for cart transfer. If you do not qualify, forward the developer docs to whoever owns your checkout stack and put UCP version 2026-04-08 on the roadmap before Black Friday planning locks.

Search Engine Journal: Google UCP update lets merchants enable cart transfer to siteGoogle Merchant Center help: UCP checkoutGoogle developer docs: UCP Cart overview

Google's holiday agentic push rewards feeds, punishes pages

In the same holiday post, Google made AI performance insights in Merchant Center generally available in the U.S., Canada, Australia, India, and New Zealand, opened a Business Agent beta inside YouTube ads, and expanded loyalty-program data ingestion. Google says merchants who adopt its core feed best practices see an average five percent conversion lift the following month. In testing with Lululemon, conversational attributes from the retailer were incorporated in relevant AI Mode product recommendations half the time.

This hits every merchant whose product feed has been treated as a low-priority technical asset. Agentic surfaces read the feed, not your product pages. Stale titles, missing sizes, empty conversational attributes, or thin descriptions now cost you placement in AI Mode and Gemini shopping answers during the highest-intent weeks of the year.

The move this week: pick your top 20 revenue SKUs from last holiday, open their feed entries, and fill every optional attribute Google exposes, including conversational descriptors like use case, occasion, and material feel. One hour of feed work on the SKUs that already sell beats a week of feed work spread across the catalog.

Google blog: Boost your holiday sales with agentic commerce updates

Cloudflare's Bot Preference Sync is now the AI crawler control surface

Cloudflare's Bot Preference Sync, published on August 21, generates your robots.txt from the AI crawler policy set in the dashboard. It works on every plan from Free to Enterprise, and it writes rules per category (Search, Agent, Training) rather than per crawler. It prepends generated rules to the existing file inside BEGIN and END comment markers. This sits directly in front of Cloudflare's September 15 change, which flipped the default position on AI crawlers for sites it serves. New domains that declare ad monetisation now get Training set to Disallow by default.

This hits anyone whose site is behind Cloudflare and who has not opened the bot settings since mid-September. Two failure modes are live right now. You may be blocking training crawlers you actually want indexing you for AI citations, or you may be handing your content to scrapers you meant to block. Either state is a licensing and citation revenue mistake.

The move this week: log into Cloudflare, open the AI bot settings, confirm the Search, Agent, and Training categories match the policy you would actually defend in writing, and turn Bot Preference Sync on. Then check your live robots.txt to confirm the generated block appears above your existing rules.

Cloudflare blog: Introducing Bot Preference SyncSearch Engine Journal: Cloudflare will write your robots.txt

Google Search Profiles drop to 10,000 followers

Google Search PM Ibrahim Badr announced on September 16 that Search Profile eligibility now sits at 10,000 followers, down from 35,000 in August and 100,000 at launch. The count is per account on any one of YouTube, Instagram, X, or TikTok. Profiles remain U.S.-only. The update also adds multi-profile management from a single login and an improved article display. Google did not announce a ranking boost.

This hits U.S. publishers, creators, and multi-brand operators who cleared 10,000 followers on any single platform and do not yet have a Search Profile claimed. The profile is a direct follower-acquisition and presentation surface on the SERP. Not claiming it is leaving a Google-owned real-estate slot unfilled while a competitor can still take it.

The move this week: check your follower counts across the four eligible platforms, and if any account is over 10,000 claim the profile now. If you run several brands, use the new single-login multi-profile setup so one operator holds the keys rather than distributing them across multiple accounts.

Search Engine Journal: Google Search Profiles now available at 10,000 followers

Pinterest opens Visual Search Ads in beta

Pinterest announced Visual Search Ads and an AI Restyle feature at Pinterest Presents on September 17. Visual Search Ads place advertisers inside Pinterest search results and Pin close-ups, with a bid for the top response and a click through to the advertiser's own site. Pinterest reports more than 80 billion searches per month, most of them visual, and Q2 2026 revenue of 1.18 billion dollars, up 18 percent year over year, on 640 million monthly actives. The Visual Search Ads beta rolls out in selected markets over the coming weeks. Restyle launches in the U.S. and Canada.

This hits home decor, fashion, furniture, and retail advertisers whose products photograph well. Early beta inventory allows advertisers to gain an advantage before CPMs rise.

The move this week: open Pinterest Ads Manager, check whether Visual Search Ads access is available to your account, and if yes ship one small test campaign against a hero product with existing high-performing organic Pins. If you do not have access, email your Pinterest rep or partner today and ask to be added to the beta list.

Pinterest Newsroom: New Pinterest Visual Search and Performance AdsSocial Media Today: Pinterest announces visual search ads and an AI restyle feature

Mueller reframes a canonical de-indexing scare as an error-shell problem

John Mueller responded to a widely shared report of a site being de-indexed and replaced by an unrelated casino domain as its canonical. Mueller said the likely cause is not a hijacked canonical but a site serving a generic client-side error shell to Googlebot, which then treats every URL that renders the same error as duplicates of each other. The fix is the render, not the canonical. His recommended tools are Search Console live URL inspection, automated pre-launch render tests, and scheduled monitoring on critical pages.

This hits publishers and ecommerce operators running JavaScript-rendered or dynamically-served pages, particularly anyone who has been chasing a canonical fix while the underlying pages return an error shell to the bot. A real cross-domain canonical requires the tag on your own domain, so if it is not there this is not a canonical problem at all.

The move this week: pick your five highest-revenue URLs, run each through Search Console's live URL inspection, and confirm the rendered HTML contains the actual product or article content, not a loading state or error page. If it does not, that is the ticket for your dev team today, ahead of any canonical review.

Search Engine Journal: Google responds to cross-domain canonical de-indexing report

EEA local businesses added to Google's new supplier unit

Google updated its Search Central documentation on September 18 to add local business queries to the two EEA-only result units it rolled out on September 8. The aggregator unit shows listings from approved vertical search services, with the top-ranked provider expanded by default and only one aggregator unit shown at a time. The supplier unit sits alongside and shows the businesses themselves. Brick-and-mortar businesses and service providers like plumbers are eligible for the supplier unit, and Google says no data beyond what it can already crawl is required.

This hits EEA-based local operators across every vertical now covered: hotels, flights, long-distance transport, products, and local services. If your Google Business Profile category is wrong or your site does not clearly state what you do and where, you risk losing clicks to directory competitors who appear in the same unit.

The move this week: open your Google Business Profile, verify the primary category is the tightest accurate match, and check that your homepage headline and footer name the service and the city in plain text. Then search your top three service queries from an EEA IP and see whether you appear in the supplier unit or whether a directory competitor is taking the click.

Search Engine Journal: Google adds local businesses to EEA search result unitsGoogle Search Central: Supplier unit documentationGoogle Search Central: Aggregator features documentation

Checked, and quiet

LinkedIn's network tab test and post proofreader launch, Reddit's SMB holiday guide, X's promotion guidance, and YouTube's channel pivoting advice all landed this week but do not change what a small marketing team should do on Monday. Nothing in that batch moves money on its own. We are watching Meta's Muse agent expansion and Google's Discover topic overviews, but neither has reached the point where a specific operator gains or loses on the basis of this week's news.

Signals checked weekly, quiet weeks admitted. If you want these same signals checked against your own site and stack instead of a general read, run the free scan at app.signalandscience.com and we will point you at the ones that actually apply.

Get the next Signal & Science brief when there is something worth acting on.