Platform Digest: Google, Meta & ChatGPT Ads, Sept 8–15

Three separate signals this week — none of them routine. Google's organic rankings quietly lurched and then partially corrected themselves without any formal explanation. Meta changed what the word "engagement" actually means in your reports. And OpenAI rewrote the rules of who can buy ads on ChatGPT, without a public announcement. Each story stands on its own, but together they point to the same thing: platforms are tightening control in ways that are genuinely easy to miss until your numbers move.

Google: A Ranking Blip Nobody Announced — and a New Bidding Beta That's Worth Watching

Google confirmed nothing publicly this week — which is itself the story. Search Engine Roundtable reported on September 14 that there was a definite ranking shift around September 4, affecting a range of sites tracked by SEO analysts Glenn Gabe and Marie Haynes. The stranger part: in many cases, whatever moved on September 4 had partially reverted by September 13. Google has acknowledged nothing and given no indication whether this was an intentional update, a rollout that got pulled, or something else entirely.

For paid search advertisers, the organic volatility is background noise — but it matters if you run campaigns whose Quality Scores or landing page relevance signals are influenced by organic performance. A site that was algorithmically suppressed for nine days and has since partially recovered may show performance anomalies in ad accounts that have nothing to do with your bids or creative. If your campaigns showed unusual cost-per-click or impression share movement between September 4 and 13, annotate it now. Wait until attribution reports surface it and it'll look like a bidding problem.

Separately, a more actionable development: Google Ads this week surfaced a beta called Product Value Optimization. Spotted by advertiser Adriaan Dekker and reported by Search Engine Roundtable, the feature lets advertisers configure value adjustments to guide automated Smart Bidding toward products that matter most for their business goals — whether that's margin, seasonal sell-through, or bestseller velocity. You can apply adjustments by brand, category, or other product attributes, and the system optimises delivery outcomes accordingly.

For ecommerce accounts running Shopping or Performance Max campaigns, this is genuinely useful. Most advertisers running PMax today feed the algorithm a flat ROAS target and then complain it over-indexes on low-margin or easy-to-convert SKUs. Product Value Optimization gives a way to correct that without manually splitting campaigns by product tier — no workaround required, just cleaner intent signals going in. It's in beta, so not all accounts will see it, but if you manage catalogue-heavy campaigns for retail or D2C clients, check your campaign creation flow now.

Also: September is the month Google's AI Max auto-upgrades are in active rollout. Campaigns using Dynamic Search Ads, automatically created assets, and campaign-level broad match are being migrated to AI Max without requiring any action from advertisers. Google expects all upgrades to conclude by end of September. If you haven't audited which campaigns are in the migration path and set a clean pre-migration performance baseline, do it before the end of this week — you'll want a reference point to tell the difference between a platform effect and a genuine performance shift.

Meta: "Engagement" No Longer Means What It Used To

Meta replaced the Post Engagement conversion type in Ads Manager with a new metric called Interactions — and this change is already confirmed live across accounts. On the surface it reads like a rename. Three things changed in how the metric is defined and tracked.

First, clicks no longer count as engagement. If your engagement campaigns were previously pulling click data into the Interactions column — whether intentionally or as a side effect of how you tracked performance — that data is now gone from the metric. Second, engaged-view attribution is now called "engage-through" attribution, and it's been extended to apply across every ad format, not just video. Third, every new ad set targeting the old engagement objective will automatically optimise toward the new Maximize Interactions performance goal, while existing ad sets remain on Post Engagement until you change them manually.

The practical consequence is immediate for anyone running monthly reports. Engagement numbers will look lower in September than in August — not because performance dropped, but because the definition narrowed. Any automated rules inside Ads Manager that read engagement counts or fire at engagement thresholds are now reading different numbers. Check them this week.

There's also a meaningful strategic shift embedded in this. Meta has been moving toward treating creative quality as the primary variable in ad performance, and the Interactions update reinforces that direction hard. Clicks, which could previously mask poor creative through accidental or curious engagement, no longer pad the metric. What remains in the Interactions count is intentional, non-click engagement — reactions, comments, shares, saves. For a real estate developer running awareness campaigns on Facebook, a lower Interactions count this month likely means the creative wasn't prompting the kind of deliberate response that predicts downstream intent. The number is more honest than the one it replaced.

Running alongside this: Meta's placement exclusion controls at the ad set level are also being phased out, replaced with bid value rules that can reduce bids by up to 90% on unwanted placements but cannot turn a placement fully off. Account-level Placement Controls remain available under Advertising Settings — but that lever applies to an entire account, not individual campaigns. If you manage multiple client brands from one account, as many agencies do, that single toggle is too blunt to be useful for brand safety exclusions. Document every exclusion currently set at the ad set level, build matching bid-reduction value rules now, and plan for the checkboxes to disappear. There's no specific deadline date on record.

OpenAI: ChatGPT Shuts Out Competitors — Without Saying So

This was the cleanest story of the week and the most consequential for how advertisers should think about the ChatGPT ad platform. On September 9, The Information reported that OpenAI had quietly updated its advertising policies to block rival AI companies from promoting image- and audio-generating products inside ChatGPT. The change was never publicly announced. It came to light when Adobe's Senior Director for Americas Media, Doug Wyatt, confirmed that OpenAI had notified advertising partners that campaigns for standalone image and voice generation tools would no longer be approved.

Adobe found out the hard way. The company had been part of OpenAI's early advertising pilot since February, running campaigns for both Acrobat Studio and its Firefly AI image generator. Firefly campaigns are now blocked. Wyatt described the situation plainly: "Image and voice/audio are closed off. Video's still open…for now."

The timing is deliberate. OpenAI released ChatGPT Images 2.5 and its ChatGPT Live voice product around the same window it tightened the policy — exactly the two categories it just locked competitors out of. The pattern is recognisable: a platform expands into a product category, then restricts paid promotion of competing products in that same category. Broadcast networks and streaming platforms do this routinely. What makes this different is scale and opacity. ChatGPT has over a billion monthly users. The policy change wasn't reflected in any published advertising guidelines. And it extended beyond paid ads — ChatGPT simultaneously stopped returning outbound citations and external links for image-related queries, so organic discovery of rival tools has been narrowed too.

OpenAI is chasing a $2.4 billion advertising revenue target for 2026, having already crossed a $1 billion annualised run rate within roughly 200 days of launch. The platform reached self-serve availability across India, Europe, the Middle East, and North Africa in late August, meaning Indian advertisers can now buy ChatGPT inventory directly through Ads Manager without going through a sales team. That access is real and the audience is valuable. But this week confirmed something that won't get less true over time: ChatGPT is a platform run by a company that competes directly in many of the same categories its advertisers operate in. The competitive exposure will only grow as OpenAI launches more products.

For most brands — in real estate, ecommerce, BFSI, and other sectors — the ad ban on generative AI tools doesn't affect their campaigns directly. But it signals clearly that ChatGPT's ad inventory will be shaped around OpenAI's own commercial priorities, not just advertiser demand. Categories that feel safe today may close without warning. Anyone building significant spend into ChatGPT Ads should diversify placements and monitor category adjacency carefully.

The Week in Brief

Three platforms, three different flavours of reduced advertiser control: Google moving campaigns to AI Max without a switch to flip back, Meta replacing placement exclusions with bid discounts and redefining core engagement metrics, and OpenAI blocking competitors from categories it wants for itself. The common thread is platforms optimising for their own systems and revenue models in ways that are easy to miss until performance data reveals them weeks later.

Q4 is close enough that any of these changes — the AI Max migration, the Interactions metric shift, the evolving ChatGPT policy — can quietly reshape how your campaigns perform before you've had time to run a proper audit. If your paid media accounts span Google, Meta, and emerging AI channels and you want a clear-eyed look at where your setup stands heading into the quarter, Transformics is the right conversation to have. Start it at transformics.co.in.