To me, Prebid Ascent: London 2026 felt like a practical conversation about publisher control. While sessions dove into the usual topics of AI, mobile monetisation, transparency, privacy, and Google’s remedies, the thread running throughout was that publishers still want more say over how their inventory is packaged, valued, sold, and measured.

Focus AI on packaging your supply

The opening sessions made it clear that Prebid’s mission still starts with publishers, even as agencies, buyers, and holding companies become more involved in shaping what comes next. Anne Coghlan, Co-Founder and COO of Scope3, captured that neatly by comparing a global CMO with a small business owner trying to sell cat mugs. Different scale, different budgets, and different tools, but the same basic need - be discoverable, buy more easily, and understand what is working.

That’s where the agentic AI discussion got interesting. We know that AI won’t magically fix programmatic, but standards like the AdCP could make premium inventory easier to discover and transact, combining some of the control of direct sales with the scale of programmatic. The consensus is that testing now is smarter than waiting for a perfect winner to emerge - and it’s good to see that some early campaigns are already live.

There was also a useful challenge from the publisher side. Some of today’s agentic standards still make the seller too passive, when really, publishers need better ways to package inventory, expose context, and show why one impression is not the same as another. Paul Farrow, Principal Product Manager at Microsoft, took this further by focusing on the information layer itself - machine-readable content, licensing, attribution, measurement, reporting, and dispute resolution. 

The governance point was just as strong, with speakers arguing for embedded human judgement, clear audit trails, and rules that can be enforced before anything goes wrong.

Close the gap between engineering and adops

The mobile header bidding panel was refreshingly honest. Waterfall bidding is still here - and panellists from Moloco and Altitude know that better than most. Hybrid setups are the reality for most publishers right now, while gaming publishers have been ahead of the curve for a while, and non-gaming publishers are catching up. What's sharpening that catch-up is AI chipping away at web traffic, pushing more publishers to treat apps as a serious revenue stream rather than a side project.

The friction here isn't about whether header bidding is a good idea, because most people agree it is. It's about the gap between engineering teams focused on making apps work and adops teams who think in line items and price granularity. Those two worlds haven't fully converged on mobile the way they have on the web, and that gap keeps slowing adoption down. Add SDK sprawl and latency pressures, and it's clear why publishers keep falling back on what they know. 

When it comes to identity, deterministic IDs still have value but they don't scale easily, so publishers and partners are spending more time on probabilistic approaches, better standards, and more useful signal handling.

Take control of the final auction

The sell-side decisioning session took that same practical tone into the auction itself. The focus was on log-level data, machine learning, prebid modules, real-time inputs, and controlled testing at the point where publishers still have room to act. 

One case study described an average of around 9% RPM lifts over six months, by adjusting decisioning at ad-opportunity level rather than relying on static rules. That kind of consistency shows that publishers have more room to act in the final auction than they tend to use, but capturing it requires the data infrastructure and testing discipline to make it real. This gives publishers a way to optimise the final auction with more precision, instead of leaving every fast decision to someone else upstream.

Trace the dollar flow, not the talking points

The transparency and ‘tech tax’ panel had more edge to it than these conversations usually do. Scott Siegler from Amazon Publisher Services made the point that lower price points don't automatically equal better performance: you can get an impression dirt cheap on a supply path that simply doesn't deliver. Charlstie Veith from Magnite also warned against SSPs bending seller declaration rules, with some publishers being pushed to declare ‘direct’ relationships that should properly be listed as ‘reseller’.

The Washington Post’s Jason Tollestrup, speaking from the publisher side, suspected publishers with genuinely tight rev shares might be getting throttled by partners chasing their own margin targets, and called for a more transparent middle of the chain. The practical advice: trace the dollar flow, get into log-level data, compare supply paths properly, and don't sacrifice optionality for the sake of simplicity.

Meanwhile, Alan Chapell, President of Chapell & Associates, had already set some of that context earlier in the day, widening that conversation beyond supply paths. His argument was that privacy, competition, copyright, AI, and platform power now move together, and publishers should stop treating them as separate policy buckets. He was also blunt about consent fatigue, and pointed out that purely contextual monetisation is too weak on its own for most publishers. The more realistic path is probably more experimentation around less intrusive, but still commercially workable, targeting and measurement.

Test the Google remedies now

Alan’s advice fed neatly into the closing session on Google’s remedies. The panel understood that a Google bid adapter or unified auction wouldn’t solve everything overnight. The definition of the ‘open web’ will decide how much practical value publishers get, especially if mobile, video, PMPs, native, and private auction activity sit outside of the frame. 

The near-term upside could still be meaningful if publishers gain more freedom to access AdX without GAM and rethink how they manage auctions, deals, and ad serving… but only if they start testing those scenarios now.

Prebid Ascent London showed that publisher control is still achievable, but it now depends on better packaging, better data, better testing, and better rules. It’s a more demanding version of control than the market asked for a few years ago, but it’s also a more durable one. From AI standards to auction logic, the fight for that control is very much still on - and the publishers who engage with it now will be better placed than those who wait. 

If you’re thinking through what that looks like in practice for your own monetisation strategy, get in touch with our team to explore how these shifts could affect your setup, and your revenue.