Jamie Roberts, founder of Loud! OOH, on how out of home ROI actually gets proven, and why AI is quietly breaking the part of the chain nobody owns.
Nobody in out of home can tell you what a poster is worth by looking at the poster. It is the first thing you learn in this job and the thing the industry is least keen to say out loud. The board is not the measurement. What somebody does in the hour after walking past it is the measurement.
That is not a weakness. It is simply how the channel has always been evidenced. The trouble is that the evidence now runs through something none of us control.
How out of home ROI actually gets proven
Route puts out of home’s weekly reach at 98% of the UK population. Outsmart’s OutPerform study with Ipsos found exposure drives a 17% uplift in smartphone brand actions, rising to 38% across the strongest campaigns, and that 66% of those actions go direct to the brand: a search for the name, a visit to the site. Rapport’s research puts the search uplift at 129%.
Read those back and notice what they share. Every one of them measures out of home by what happens in search afterwards. Brand search volume, site sessions, footfall modelled off device movement, promo codes, regional holdouts. It is a decent toolkit. Every part of it sits downstream of somebody looking something up.
Which is a problem, because looking things up has changed
Pew Research tracked 68,879 Google searches across 900 adults. When an AI summary appears at the top of the page, users click a traditional result 8% of the time, against 15% when no summary appears. Clicks on the links inside the summary itself: 1%.
Previsible analysed 6.77 million AI-referred sessions across 166 GA4 properties and found the channel grew 9.9 times between November 2024 and May 2026, with ChatGPT responsible for 92.4% of it. Similarweb puts the conversion rate on those referrals at 7.1%, behind paid search and ahead of everything else.
I am not going to oversell it. AI referrals still account for somewhere between 0.1% and 2.8% of total site traffic depending on whose study you read. It is a trickle. But here is the part that ought to bother anyone selling this channel. If your poster prompts a search, and that search returns an answer assembled without you in it, the campaign still worked. You just cannot prove it, and neither can your client. That is not an attribution gap. That is an attribution model failing quietly while everybody carries on reporting off it.
So we published every price we charge
Out of home has just had its best year on record, £1.44bn in 2025, up 2.6%, with digital now 67% of the total according to Outsmart and PwC. Also a market where you can ring five suppliers about the same panel and get five different numbers, none of them written down anywhere. A model cannot cite a number that only exists in a phone call.
Digital had this argument a decade ago and lost it in public. When ISBA and PwC first traced programmatic spend end to end in 2020, 15% of advertiser money could not be attributed to anybody at all, and only around half of it reached publishers. The follow-up study pulled that delta down to 3% and lifted publisher share to 65%. The money did not change. The visibility did.
In late June we booked two adjacent 48-sheets on York Street in Leeds. One carried a £650 quote a client had been sent. The other carried our published 48-sheet rate for the same format, £325. We named nobody. New website traffic rose 188%, five press placements followed, and client campaigns were booked off the back of it. We put the whole breakdown on the site afterwards, including the parts that did not work.
Because the boards were a stunt, and stunts decay. The traffic spike flattened inside a fortnight. What did not flatten was the pricing page, because it goes on answering the same question every day long after the coverage dies.
Building a site for a reader that will never look at it
Four things, none of them glamorous. Prices as structured, extractable text rather than image files or JavaScript counters that render as zeros in the source. Direct answers inside the first two sentences of a page, because that is the part that gets lifted. The same figures repeated identically everywhere we appear, since inconsistency is what makes a model hedge. And an acceptance that most citations come from third parties rather than your own domain, so earning coverage beats tinkering with your own markup.
As far as I can establish, no other UK out of home agency has done this deliberately. I would rather be wrong about that. An industry in which exactly one agency is legible to AI is not a competitive advantage worth having. It is a warning.
The uncomfortable bit
You cannot make a website AI-readable if it has nothing to say. Schema markup on a page that reads “contact us for a quote” gives you a beautifully structured way of saying nothing. Our sector’s problem was never technical. It is that much of the pricing model depends on the buyer being unable to compare, and that model has a short runway left.
For northern SMEs this is not abstract. If the only out of home numbers a model can find are enterprise ones, it will keep telling small businesses that billboards are not for them. That is a self-fulfilling forecast, and we built it ourselves, one “price on application” at a time.
Jamie Roberts is founder of Loud! OOH, an independent out of home advertising agency with offices in Leeds and London.
SOURCES (for editorial verification, not for publication)
Weekly reach 98%, search uplift 129%, talkability: Outsmart, The Superpowers of Out-of-Home, March 2026 (sources cited: Route, Rapport)
Smartphone brand actions 17% / 38% / 66% direct to brand: Outsmart OutPerform, with Ipsos and Route
https://www.outsmart.org.uk/resources/outperform
UK OOH revenue £1.44bn 2025, +2.6%, digital 67%: Outsmart / PwC, February 2026
Programmatic unknown delta 15% to 3%, publisher share to 65%: ISBA / PwC Programmatic Supply Chain Transparency Study
https://www.isba.org.uk/knowledge/second-programmatic-supply-chain-transparency-study
Click rates 8% vs 15%, 1% on in-summary links, 68,879 searches: Pew Research Center, July 2025
AI referral growth 9.9x, ChatGPT 92.4% share, 6.77m sessions: Previsible AI Traffic Report, 2026
https://previsible.com/seo-strategy/ai-traffic-report-july-2026
ChatGPT referral conversion 7.1% vs paid search 7.8%: Similarweb clickstream, April to May 2026
https://www.similarweb.com/blog/marketing/geo/gen-ai-stats
AI referrals 0.1% to 2.8% of total traffic: SearchSignal benchmark, 2026
https://searchsignal.online/research/ai-search-referrals-citations-2026
York Street campaign results, 188% and five placements: loudooh.co.uk/case-studies/york-street-billboard-leeds/
