For much of the last decade, premium publishers could run an inefficient ad stack without really feeling its impact. Any loss through inefficiencies was masked by traffic growth. If organic volumes grew 8% and the programmatic stack leaked 3%, the numbers still moved in the right direction and it didn’t matter how much 3rd party tech you had sellotaped together.
But with Google’s AI Overview (AIO) functionality essentially decimating organic traffic for whole sectors of publishers, that calculation has had to change pretty quickly.
Organic search traffic across premium content publishing is negative, year on year. AI Overviews are taking clicks that used to land on publisher pages, particularly on informational queries, and the publishers most exposed are the ones whose audience strategies were built around search visibility. Several major UK publishers have said as much publicly, framing the current moment as a period of quantifying impact rather than questioning whether it exists.
Some have offset the decline with growth from Google Discover, the algorithmically curated mobile feed that now accounts for roughly two thirds of Google-referred traffic to news publishers. In practice, Discover traffic is volatile (entire publisher feeds have been wiped overnight by core algorithm updates), low-loyalty (readers are less likely to subscribe or explore the site) and outside the publisher’s control.
The implication is that publishers carrying inefficiencies they had at higher traffic volumes are now losing ground in absolute terms. Yield leakage that was tolerable before might not be as easy to ignore today.
THE INEFFICIENCIES THAT GROWTH WAS ABSORBING
While every publisher setup is unique, here are three categories of waste appear that we see appearing consistently in publisher stacks.
The first is unviewable refresh impressions. Most refresh configurations run on static time intervals (typically 30 seconds) regardless of whether the ad unit is in the viewport. The result is a steady output of impressions that were never seen. Buyers are increasingly filtering on viewability signals. A refresh impression that never entered the viewport degrades the quality signal attached to every other impression on the page. View-based refresh logic (triggering only when a unit is in-view and has met a minimum engagement threshold) is the fix and the buyers who matter most are already rewarding publishers who implement it.
The second is bid redundancy. Two common forms: sending all bid responses to the ad server rather than the winning bid only, adding decisioning overhead and latency for no yield benefit; multi-size placements generating separate bid requests per size, multiplying queries per second without corresponding demand.
Both patterns degrade perceived supply quality at the DSP and SSP level. The Trade Desk, PubMatic and Magnite are all actively optimising supply paths and following a bid shading strategy. A publisher generating unnecessary QPS load is not what the buyers want in their auctions.
The third is dead identity weight. Publishers carrying multiple identity modules on-page, of which only one or two are returning live match rates. The rest make calls, get nothing back and add page weight for zero targeting uplift. The CPM delta between identified and unidentified inventory has widened steadily as third-party cookie deprecation has progressed. Every identity module on the page should either be contributing live matches or be removed entirely. Carrying dormant modules is a double cost (latency paid, addressability not received) and there is no traffic level at which that trade-off makes sense.
HOW THESE COMPOUND WHEN AIO TRAFFIC DOWN?
Individually, none of these are ideal, but together, they compound in a way that is worse than the sum of the parts.
Buyers aren’t evaluating impression price in isolation. Supply path quality, QPS load, viewability signals, identity match rates, ad load latency. These all feed into how DSPs and SSPs score and prioritise inventory. A publisher running fewer impressions through a degraded supply path gets hit on both sides. Lower volume and lower value per unit. Beyond reducing what’s you are physically able to sell, the loss of search traffic makes the cost of every existing inefficiency more visible, because each impression carries more weight.
The reason programmatic efficiency matters more now than it did three years ago is that the growth which was offsetting them has stopped.
ACTIONABLE RECOMMENDATIONS
View-based refresh logic, not time-based. Look at your bid efficiency. Ensure winning bids only to the ad server, bundled multi-size requests rather than duplicated ones. Identity modules that return live match rates only, remove dormant or redundant ID modules. Full supply chain transparency through properly maintained sellers.json and ads.txt. And incremental demand integration from sources that bring genuinely differentiated buyer pools rather than duplicating what’s already in the auction.