The Real Reason Meta Ad Performance Is Degrading

Why you're getting this: You've crossed paths with CE Digital somewhere along the way. Every week we share what we're actually seeing inside high-spend ad accounts, plus how we're using AI and our internal data tool, AdSignl, to analyze performance at scale. Short, practical, no filler.

Not for you? No problem. Unsubscribe any time.

Alright, let’s get into it:

If you run paid media on Meta and you've been quietly wondering whether something is broken across every account, not just yours, this piece is for you.

I've spent the last several weeks pulling together everything Meta has disclosed in its SEC filings, what executives have said on earnings calls, what Reuters uncovered in leaked internal documents, what the California Attorney General alleged in court filings, and what independent traffic measurement firms reported through mid-2026. The pattern is clear enough that I want to put it on paper.

The short version: Meta's ad auction in 2026 is operating under three structural pressures that compound on each other. They are documented in Meta's own filings. None of them is your creative. And clients deserve to hear it framed honestly.

On April 29, Meta reported Q1 2026 earnings. Three numbers tell the story:

Read that again. Meta delivered 19% more ads to a user base that grew only 4%. That gap means every existing Meta user is being shown roughly 14 to 15 percent more advertising than they were a year ago. CPMs climbed 12% on top of that volume.

Across the broader market, the DTC ecommerce CPM picture is even clearer. Industry benchmark data shows average Meta CPMs for ecommerce climbed from roughly $11.82 in 2025 to $14.19 in 2026, a 20% year-over-year increase. Healthcare verticals saw CPMs rise from $22.76 in January 2025 to $38.70 in January 2026, a 70% jump.

A 20% CPM increase on a platform whose user base grew 4% is not normal demand-supply pricing. It is the visible consequence of three structural problems Meta has chosen to manage rather than solve.

Meta says this themselves. From the 2025 10-K filed with the SEC: "The increase in average price per ad in 2025 was driven by an increase in advertising demand... This increase was partially offset by a higher number of ad impressions delivered, especially in geographies and in products, such as Reels, that monetize at lower rates."

That single sentence is Meta admitting two things at once. Advertiser demand keeps rising. But the new ad supply they're adding to absorb that demand is structurally lower-quality: Reels, Asia-Pacific, lower-monetizing surfaces. The high-yield inventory, Facebook Feed and Instagram Feed, is not where impression growth is coming from.

CFO Susan Li named the underlying mechanic in 2023 and the language has carried through every quarter since: "There are structural supply constraints with the Reels format as people view a Reel for a longer time than a piece of Feed or Stories content, which results in fewer opportunities to serve ads in between posts. That will likely make it more challenging to close the monetization efficiency gap than it was with Stories."

Reels now accounts for 50% of time spent on Instagram. Video time spent on Instagram was up more than 30% year over year in Q4 2025. As user attention shifts from Feed to Reels, the dollar value per minute of user attention drops because each minute generates fewer ad slots.

The most underreported admission came on the Q4 2025 earnings call. From CFO Susan Li: "During the second half of 2025, Meta's initiatives to redistribute ads across users and sessions on Facebook led to a nearly four times larger revenue impact than increasing ad load on Facebook."

Read carefully. Ad load on Facebook has reached a point where adding more ads degrades user experience faster than it generates revenue. So Meta pivoted to algorithmically deciding which users should see more ads in a given session. The 4x multiplier means redistribution is now the dominant revenue lever, not ad load increases. Prime inventory is full.

The capex tell makes the same point. Meta raised 2026 capex guidance to $125 to $145 billion, roughly double 2025's $72 billion. That spend is going to AI infrastructure that does two things: improve ranking efficiency so each impression yields more revenue, and generate AI content (Vibes, Muse, Meta AI) to fill the inventory gap human creators are no longer filling. If the underlying supply problem were not large, $135 billion in compute would not be the answer.

Source: Meta SEC 8-K filings, Q4 2024 through Q1 2026

The auction does not run in a vacuum. Ads sit inside a feed of organic content. When the organic layer degrades, the algorithm's signals degrade with it.

Adam Mosseri, head of Instagram, said something significant in January 2026. Meta would push creators toward Meta Verified specifically so the platform could rank their content higher, because verification was the way Meta could confirm content came from actual humans. That's an admission that algorithmic distinction between human and AI content has broken down. Instagram needs creators to pay for the privilege of being identified as not a bot.

The publisher data confirms what Mosseri is responding to. Per Digiday's research, 86% of publisher professionals posted content on Instagram in Q3 2024, down from 91% a year prior. Publishers reduced daily Instagram posting from 69% to 52% in twelve months. Only 38% of publishers consider Instagram valuable for revenue generation, down from 47%, with 22% now calling it not very valuable, up from 7%.

When the economics stop working for publishers and creators, they post less. Less posting means less organic content. Less organic content means the algorithm has fewer signals to work with, and the inventory pool shrinks at the only surfaces that monetize well.

The engagement data shows the cumulative effect. Buffer's 2025 cross-platform engagement report found Instagram's median engagement rate dropped from approximately 7.3% in 2024 to 5.4% in 2025, a 26% year-over-year decline. Socialinsider's Q1 2026 data shows Reels engagement at 0.50% (down from 0.52% in Q4 2025), single-image posts at 0.35%, and carousels at 0.52%. Business accounts have been hit harder than personal accounts, with reach down 47% versus 21% for creator accounts.

Facebook organic reach, per Social Status benchmarks, sat between 1% and 2% across 2024 and 2025, down from 16% in 2012.

Meta's response has been to substitute AI content for human content. In September 2025 they launched Vibes, an entirely AI-generated short-form video feed that cross-posts into Instagram and Facebook Reels. In a Q3 2025 earnings call comment, Mark Zuckerberg framed this strategy explicitly: AI-generated video "will help increase the amount of content inventory that can be shown in Instagram and Facebook, and therefore, should enable an increase in engagement there."

Translated: Meta is using AI content as supply expansion for the advertising auction because human creator output is not scaling fast enough.

Recent academic research from the University of Florida, published March 2026, names the second-order consequence: "Now there is a flood of relatively low-quality content. Because the quantity is so large, it congests the recommendation systems, so it gets harder to encounter the truly high-quality content."

The recommendation system that serves your ads is the same one that serves organic content. When it is congested with synthetic content, both systems degrade together.

This is the most damaging part of the picture and the one most paid media operators don't realize is documented in court filings.

Start with what's measurable. Imperva's 2026 Bad Bot Report, released April 29, 2026, found that automated bot traffic accounted for more than 53% of all web traffic in 2025, up from 51% in 2024. AI-driven attacks rose 12.5x year over year, from 2 million blocked daily incidents to 25 million. On June 3, 2026, Cloudflare CEO Matthew Prince shared Cloudflare Radar data showing bots now generate 57.5% of HTML web traffic, with humans at 42.5%.

Humans are now the minority of internet traffic across two different measurement methodologies. Both moved past the 50% threshold within the last 24 months.

For Instagram specifically, a September 2025 report from venture capital firm Galaxy Interactive estimated as many as 95 million accounts (9.5% of total) could be fake or automated.

That sets the stage for what Reuters uncovered. Across November and December 2025, Reuters published an investigation based on Meta's internal documents. The findings:

Meta projected approximately $16 billion (10% of 2024 revenue) would come from advertisements promoting scams or banned goods. Meta's platforms show users approximately 15 billion "higher risk" scam advertisements daily. Meta earns roughly $7 billion in annualized revenue from this single higher-risk category.

What Meta does with suspected scammers is the part that directly affects your CPMs. Per the Reuters reporting: "To advertise on Meta's platforms, a business has to compete in an online auction. Before the bidding, the company's automated systems calculate the odds that an advertiser is engaged in fraud. Under Meta's new policy, likely scammers who fall below Meta's threshold for removal would have to pay more to win an auction. Documents from last summer called such 'penalty bids' a centerpiece of Meta's efforts to reduce scams."

The threshold for outright removal is 95% certainty of fraud. Anything below that, including advertisers Meta's own systems identify as up to 94% likely to be scammers, stays in the auction at inflated bid prices.

On May 11, 2026, the Santa Clara County Counsel filed a civil enforcement action against Meta in Superior Court of California. The complaint frames the consequence for legitimate advertisers directly: "The resulting flood of competitors, many of which are scam advertisers, drives up prices for legitimate advertisers. And, because the winning bidder for a particular ad placement necessarily displaces the losers, scam ads routinely displace legitimate ads... In auctions, scam advertisers drive up the financial bids placed by legitimate advertisers and therefore the price of placing legitimate ads."

The complaint further alleges that Meta "profits twice from scam advertising: first through direct advertising revenue and secondly because fraudulent advertisers drive up advertising prices for legitimate businesses competing in the same auctions."

This is no longer a theory. It is a state Attorney General allegation, supported by Meta's own internal documents, that the auction structurally inflates CPMs for legitimate advertisers.

The pile-on through 2026 is substantial. The SEC is investigating Meta's role in financial scam ads. The UK Financial Conduct Authority is investigating. The Australian Competition and Consumer Commission has filed suit. Swedish publishers (Utgivarna) sued Meta over impersonation ads. A federal class action covering advertisers harmed by inflated "potential reach" metrics was certified by Judge James Donato in the Northern District of California, with documents already showing senior Meta executives knew the metric was inflated by duplicate and fake accounts.

The final piece, and perhaps the most insidious for performance marketers, is the personalization loop. Per Reuters: "Users who click on fraudulent ads become more susceptible to seeing more of these ad types more frequently due to Meta's ad-personalization system."

Mechanically: bot accounts and vulnerable users engage with scam ads. Meta's personalization system reads engagement as purchase intent. The lookalike audiences and broad-audience pools that legitimate advertisers bid against get built partly on that corrupted signal foundation. Your Advantage+ campaign bids on users that look engaged but don't convert, because the engagement signal was generated by scam-ad-click behavior, not real shopping intent.

This is the technical mechanism by which the dead internet theory becomes a performance marketing problem rather than a philosophical one.

If you operate a DTC brand on Meta in 2026, the strategic implications are concrete.

First, CPM increases that outpace user base growth are not creative fatigue. A 20% YoY CPM jump on 4% user growth has structural causes that no amount of new creative will fix. The honest answer to a board asking why CPA went up is partly about the auction environment, not just about asset performance.

Second, the auction context for your ads has degraded measurably. When the organic content surrounding your ad is increasingly AI-generated and recommendation systems are congested with low-quality content, the algorithmic environment your ad lands in is worse than it was two years ago. Same creative, same audience, worse delivery context.

Third, the signal layer feeding the algorithm has been corrupted at a documented scale. Meta's own internal documents acknowledge $16 billion in projected violating revenue, 15 billion daily higher-risk scam impressions, and a 0.15% revenue guardrail capping enforcement spending. The personalization system that drives Advantage+ delivery is sitting on top of that signal layer.

Fourth, Meta's response to all three forces is to spend $125 to $145 billion in capex in 2026 to engineer around the problems with AI infrastructure. That tells you the company expects these structural conditions to persist, not resolve.

The implication for media planning is not "abandon Meta." Meta still has 3.56 billion daily users and remains the most efficient performance channel for most DTC brands. The implication is that the optimization assumptions that worked in 2022, when CPMs were lower and signal quality was higher, no longer hold. Diversification of paid channels, creative volume to outpace algorithmic context degradation, and stricter measurement against incremental conversion rather than platform-reported ROAS all become more important.

Most importantly, if you're being told by your agency or in-house team that your underperformance is purely a creative or offer problem, get a second opinion. There is a documented platform-level component to declining Meta performance in 2026 that is not your fault, and that no creative refresh will solve.

I would rather give clients the honest read on what is happening in the auction than tell them a comforting story. The honest read is: Meta's auction is structurally compromised in measurable ways, and the brands that perform best in 2026 will be the ones whose teams understand exactly what they're bidding into.

If you found this useful and want to talk about how this is showing up in your accounts, reply to this email or reach out at andrew@clayentertainment.com. I'm always glad to compare notes with other operators.

Andrew Clay Founder & CEO, CE Digital