Platform Digest: Google, Meta & AI News, Sept 21–28

This week the pace held. Google pushed two separate Search updates out on the same day. Meta's creator ad infrastructure went global. And Monday morning brought competing models from OpenAI and Anthropic within ninety minutes of each other — both cheaper than what came before. The through-line across all of it: every major platform is expanding capabilities while cutting costs simultaneously, which makes this a genuinely good moment to audit whether your setup is keeping up.

Google Fires Its Fourth Spam Update of 2026 — and Adds Multimodal Data to Search Console

Two distinct things happened inside Google Search on September 24, and they deserve to be read separately.

First, the spam update. Google began rolling out the September 2026 spam update that day, announcing it via Search Central on X. It applies globally across all languages. What distinguishes this one from the three spam updates earlier in 2026 — March, June, and August — is the rollout window. Each of those finished in under three days. This one carries an estimate of up to two weeks, with a potential completion date around October 8.

Google hasn't named a specific target, and no new spam policy was announced alongside it. Sites that see ranking shifts this week can't automatically pin them to any one practice — this is more plausibly a calibration of SpamBrain's detection systems than a fresh policy crackdown. Watch, don't react. Audit your content against Google's existing spam policies — scaled AI content abuse, cloaking, misleading functionality — but hold off on structural changes while day-two data is still rolling in and the update is still live.

Second, the Search Console multimodal filter. Also on September 24, Google announced that Search Console's Performance report now includes a dedicated multimodal search type, rolling out globally from that date. The filter surfaces data from Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's right-click "Search this image" feature — traffic that, according to Google's own product manager, was never counted in Web totals before. Genuinely new data, not a reclassification of existing numbers.

One caveat worth flagging: query text is unavailable for this traffic type. Because multimodal searches start with an image rather than a typed phrase, there's no keyword string to surface. You can see which pages are getting impressions and clicks from image-initiated searches, and which countries and devices are driving that activity — just not what the user was "searching for" in any conventional sense.

Open the filter now. If your site carries product images, lifestyle photography, or any visual content that shows up in Lens results, you can finally quantify that channel. Sites that have invested in image SEO — descriptive filenames, structured data, high-resolution product shots — may find meaningful traffic already flowing there that was previously invisible in their reports. Check the numbers first, then decide whether your visual content strategy justifies heavier investment.

Meta Takes Its Creator Ad Stack Global at IAB Creator Week

On September 15, at IAB Global Creator Week, Meta pushed out a substantial bundle of creator marketing updates. The headline was reach: the Creator Marketing Hub — formed earlier this year by merging the Creator Marketplace with the Partnership Ads Hub — is now rolling out globally through the end of 2026.

The specific additions carry more weight than the geographic expansion. The Hub now includes discovery filters that surface creators who have already organically featured products like yours, which is a more useful signal of fit than follower count alone. Content-level permissions now come with expiration dates, addressing a long-standing frustration around evergreen usage rights. An ad-readiness editing feature strips copyrighted music and stickers from creator content so it can run as a paid ad without a separate production step. And a one-click ad creation flow takes approved Partnership content directly into Ads Manager.

The same announcement also wired Partnership Ads into Meta's AI-agent connector, meaning brands running automated campaign workflows can pull creator content directly into that pipeline.

The practical result is a compression of the time between finding a creator and running their content as a paid ad. Brands that have been treating organic creator partnerships and paid amplification as two separate workflows — with separate clearance steps, separate editing passes, separate uploads — can now consolidate them. At any real campaign volume, that friction reduction is material. Getting a creator's post cleared, edited, and live as an ad has historically eaten time that shouldn't be eaten.

Indian brands activating regional creators at scale should pay particular attention to the discovery filters. Surfacing creators who already feature a category of product is a far more useful starting point than filtering by reach, especially for brands where authenticity of fit drives performance — which, realistically, is most of them.

OpenAI and Anthropic Both Cut Prices — On the Same Morning

Monday, September 22 produced one of the stranger moments of the year in AI releases. Anthropic launched Claude Opus 5.5 in the morning. OpenAI launched GPT-6 Sol and GPT-6 Luna roughly ninety minutes later. Both announcements led with efficiency rather than capability maximalism.

Claude Opus 5.5 is the first model in Anthropic's new 5.5 family, built for long-running agentic coding and knowledge work. It runs 40% cheaper and produces output more than 30% faster than Opus 5, while matching Claude Fable 5.1 on most tasks. API pricing comes in at $4 per million input tokens and $20 per million output — a 20% reduction versus its predecessor — with a 1 million token context window by default.

GPT-6 Sol, OpenAI's mid-tier addition to the GPT-6 generation, sits one step below the flagship Astra model that launched earlier in September. OpenAI positioned it for complex tasks and agentic workflows where cost matters as much as reasoning depth. On OpenAI's internal factuality evaluation, Sol makes roughly half as many mistakes as GPT-5.6 Sol, at a substantially lower price point. GPT-6 Luna, the smaller sibling, targets high-volume, lower-complexity tasks like summarisation and information extraction — the "clerical" tier of AI work, in OpenAI's own framing.

The ninety-minute gap between Opus 5.5 and GPT-6 Sol almost certainly wasn't coincidence. Both labs track each other's release cadence closely enough that the timing reads as deliberate. What it signals to the market is direct: the intelligence layer is getting cheaper, fast. API costs that felt prohibitive twelve months ago are being halved every few months, and that trajectory is accelerating.

For marketing teams building on top of these models — content workflows, customer service automation, lead qualification — the question has shifted. Affordability is largely solved. The real discipline now is matching model tier to task complexity. GPT-6 Luna and Claude Haiku (when its 5.5 version ships in coming weeks, per Anthropic's announcement) will be meaningfully cheaper than their flagship siblings, and for repetitive, structured tasks they may well be the smarter buy. Running a summarisation job on Opus 5.5 because it's the newest model isn't cost discipline — it's the opposite.

Both model families are available on major cloud platforms — AWS, Google Cloud, and Azure for Opus 5.5; OpenAI's API and GitHub Copilot for GPT-6 Sol and Luna — so any team already embedded in those environments can experiment with low friction.

The Week in One Frame

Google is tightening spam detection while simultaneously handing marketers better data on how users find them through visual search. Meta is collapsing the operational overhead between creator discovery and paid amplification. OpenAI and Anthropic are in a race to make frontier-level AI cheaper to run in production. Across all three, the advantage flows to teams actively checking their setups — not those who'll get to it next quarter.

At Transformics, we're running these updates through live campaigns right now: checking multimodal filter data for clients with strong visual inventory, stress-testing Meta's one-click ad flow for brands with active creator rosters, and re-evaluating model tier assignments across AI-powered lead qualification workflows for clients in real estate, ecommerce, and BFSI. If any of that sounds like work your setup needs done, it probably does.