17/03/26 // Digital Media
Meta Is Automating Your Campaigns. Here Is What That Means.
Written by: Simon
Meta is not making one change to its advertising platform. It is making several, all at once, and all pointing in the same direction: less manual control for advertisers, more decisions made by machine learning.
In the first quarter of 2026 alone, Meta has consolidated detailed targeting options, deprecated legacy campaign APIs, rolled out its Andromeda ad delivery engine globally, changed how it counts click-through conversions, and announced a new 2% location fee for UK ad delivery. Each of these would be worth paying attention to individually. Together, they represent a fundamental shift in how the platform works.
For advertisers and the agencies managing their budgets, the question is not whether to accept Meta’s automation. It is how to work within it without losing sight of what is actually driving performance.
What Meta is actually changing
These changes are not cosmetic. They affect targeting, campaign setup, ad delivery, attribution, and cost. Here is a timeline of what has happened or is happening now.
| Change | When | What it means |
|---|---|---|
| Detailed targeting exclusions removed | March 2025 | You can no longer exclude audiences by interest or behaviour |
| Detailed targeting options consolidated | June 2025 onwards | Granular interest categories merged into broad groups |
| Ad sets using removed options stop delivering | January 15, 2026 | Any campaign still using old targeting simply switches off |
| Legacy ASC and AAC APIs deprecated | Q1 2026 (Marketing API v25.0) | New campaign creation forced through Advantage+ structure |
| Andromeda ad delivery engine live globally | October 2025 onwards | Creative quality replaces audience targeting as the primary delivery signal |
| Attribution metric redefined | March 2026 | Only link clicks count. Likes, saves, and shares move to engage-through |
| 2% UK location fee | April 2026 | Digital Services Tax passed to advertisers on top of campaign spend |
Meta has been moving in this direction since the iOS 14 privacy changes in 2021 forced a rethink of signal-dependent targeting. But 2026 is when the remaining manual controls are being actively retired rather than simply deprioritised.
Where advertisers lose control
Meta’s headline framing is that automation improves performance. Internal data shows Advantage+ campaigns delivering a 22% increase in ROAS compared to manual setups, according to Meta’s own benchmarks. That number is real, but it does not tell the whole story.
Advertisers are not necessarily losing performance. They are losing visibility and control.
| What you had | What you now get | Why it matters |
|---|---|---|
| Granular interest targeting | Broad, consolidated categories | Harder to isolate niche audiences for testing |
| Interest and behaviour exclusions | Custom audience exclusions only | Less control over who does not see your ad |
| Manual campaign structure via API | Advantage+ structure enforced | Custom integrations and tools need rebuilding |
| Audience-level delivery logic | Creative-signal-based delivery (Andromeda) | Performance depends on creative quality, not audience configuration |
| Attribution counted all click types | Link clicks only; rest is engage-through | Reported click-through conversions will drop (reclassification, not decline) |
There is also a timing problem. In March 2026, advertisers across multiple industries reported sudden performance instability: fewer leads, higher acquisition costs, and inconsistent delivery. This coincided with both the Andromeda rollout reaching full scale and the attribution metric changes landing at the same time. It is difficult to diagnose what caused what when several variables changed simultaneously.
That lack of diagnostic clarity is the real cost of automation. The platform may well be making better decisions on average. But when something goes wrong, there are fewer levers to pull and less data to explain why.
Why creative now carries the weight that targeting used to
Andromeda, Meta’s ad delivery engine, is at the centre of this shift. It works in two stages. First, it scans the full ad inventory and shortlists around 1,000 candidates for each impression, based on creative signals and user intent. Second, it runs the auction to pick the winner.
Stage one is where it matters. Your ad’s creative content determines which shortlist it lands on. Manual targeting only limits which users Andromeda can consider, which means narrowing your audience can actually reduce performance by restricting the system’s ability to find good matches.
A 2025 AppsFlyer report found that 70 to 80% of Meta ad performance is now driven by creative quality rather than budget or targeting configuration. That is a significant inversion of how paid social has worked for most of the last decade.
| Old model | New model |
|---|---|
| Right audience + decent creative = results | Best creative wins the auction, audience is found by the algorithm |
| Test audiences, keep creative constant | Test creative, let the algorithm handle audience |
| Refresh creative every few months | Fresh concepts every 2 to 3 weeks |
| Creative variation = different headline or colour | Creative diversification = genuinely different ideas and angles |
This changes the economics of paid social. Creative production is no longer a supporting cost. It is the primary performance lever. Brands that treat creative as an afterthought will find their campaigns underdelivering regardless of budget.
How measurement changes too
Alongside the delivery changes, Meta has redefined how it reports conversions. Click-through attribution now counts link clicks only. Interactions like likes, saves, shares, and expanding an image to full screen are reclassified as engage-through attribution. We covered the detail of this change in our guide to Meta’s engage-through attribution.
In practice, your click-through conversion numbers will drop. For most advertisers, this is not a performance decline. It is a reclassification of where conversions are counted. But if your reporting dashboards or automated rules are built around click-through metrics, they will need updating.
| Metric | Before March 2026 | After March 2026 |
|---|---|---|
| Click-through conversions | All clicks (link, like, save, expand) | Link clicks only |
| Engage-through conversions | Did not exist as a category | Likes, saves, shares, image expands |
| Reported ROAS | Included all click types | Lower on click-through, offset by engage-through |
This is worth flagging to anyone who reviews Meta performance reports. A drop in click-through ROAS does not necessarily mean the campaign is performing worse. It means Meta is counting differently, and your benchmarks need recalibrating.
For a broader view of how attribution models compare, see our guide to marketing attribution models.
How to respond
None of this means Meta is a worse platform for advertising. For many advertisers, the automated systems will genuinely outperform what manual campaigns could achieve. But it does mean the way you get value from Meta has changed, and the skills that matter have shifted.
| Action | Why it matters now |
|---|---|
| Invest in creative volume and variety | Andromeda needs genuinely different creative concepts to optimise delivery. Colour swaps and headline tweaks are not enough. |
| Build first-party data infrastructure | Custom audiences and conversion signals from your own data are the one input Meta cannot automate away. They remain your strongest targeting lever. |
| Set up measurement outside Meta | As in-platform reporting becomes less granular, independent measurement (MMM, incrementality testing, server-side tracking) becomes essential for understanding true performance. |
| Budget for the 2% UK location fee | From April 2026, UK ad delivery costs 2% more on top of your campaign spend. For a £10,000 monthly budget, that is an extra £200 per month. |
| Recalibrate your reporting benchmarks | Click-through metrics will look different after the attribution reclassification. Update your dashboards and automated rules before drawing conclusions. |
| Keep some manual campaigns running | A hybrid approach gives you a reference point. If automated campaigns underperform, manual setups help you diagnose whether the issue is creative, audience, or platform. |
Beyond Meta, the same pattern applies. As platforms automate more of the buying process, the value of a paid social strategy shifts from configuration to interpretation. Knowing which buttons to press matters less. Knowing what the results mean, and what to do next, matters more. That is where a consultancy-led approach to media strategy earns its keep.
Is Meta removing all manual targeting?
Not entirely. Custom audiences built from your own data (customer lists, website visitors, app users) still work, and custom audience exclusions remain available. What has been removed is the ability to target or exclude by third-party interest and behaviour categories. Meta is encouraging broad targeting and letting its machine learning find the right users.
Will Advantage+ campaigns always outperform manual ones?
Not always. Meta’s own data shows a 22% average ROAS improvement, but averages hide variation. For niche audiences, regulated industries, or campaigns that need tight geographic or demographic control, a hybrid approach often performs better. The key is testing both and comparing results against your own benchmarks.
How should I adjust my Meta Ads budget for 2026?
Factor in three things. First, the 2% location fee on UK delivery from April 2026. Second, increased investment in creative production, because creative diversity is now the primary performance driver. Third, a potential need for independent measurement tools if you want visibility beyond what Meta’s own reporting provides.
If your approach to AI in performance marketing has been to wait and see, that window is closing. Meta is not offering automation as an option. It is becoming the default.
Get in touch
0148 495 9610
hello@mediaperformance.co.uk