Out of home advertising: AI buyer checks in England

What to take away

AI sits mostly in audience modelling, schedule optimisation and post-campaign attribution, not in the buying decision itself.
The audience data feeding those models is often brokered, so ICO expectations on lawful sourcing and transparency now belong in the planning conversation.
The ASA administers the CAP Code for non-broadcast OOH ads and the BCAP Code for broadcast ads; UK GDPR and the Data Protection Act 2018 apply to personal data.
Ask suppliers to name the data source, the model's purpose and the human check before a schedule is signed off.

Why AI moved into out of home planning

Out of home has always been hard to measure. A poster reaches people who never click, so planners have leaned on modelled audiences and travel data. What changed is the volume of that data and the speed of processing.

AI applications in out of home advertising cluster in three areas: predicting who passes a site, choosing which creative runs where, and estimating incremental reach. None replace the buyer. They change what a buyer can ask a supplier to evidence. That is why the wider out of home advertising trends and outlook now treats model inputs as part of the commercial conversation.

How AI applications show up in a buy

AI may enter an OOH buy in the planning layer rather than the trading layer. A media owner may run a scheduling model that shifts creative between dayparts on footfall forecasts. An agency may blend transport data with mobile location signals to model reach.

Buyers rarely see the model. They see a plan with audience numbers attached. The useful question is not whether AI was used, but which inputs produced the number. A forecast built on rail ticketing data is a different claim from one built on anonymised mobile pings.

Named examples buyers may encounter include UK media owners JCDecaux, Global and Ocean Outdoor, and programmatic DOOH platforms VIOOH, Broadsign, Hivestack and Place Exchange. These names do not establish that a particular campaign uses AI; ask the media owner or platform to identify the actual model or software, its role in the plan and who is accountable for its output.

Route is the UK's OOH audience measurement service, while JICMAIL measures mail, not OOH. For attention research, Lumen Research and Amplified Intelligence are named providers to ask about; check whether a study covers the relevant OOH format and English market.

Quividi offers computer-vision audience analytics for digital signage, a concrete example of technology relevant to DOOH. Ask whether it is actually deployed on the inventory under consideration, what it captures and how its outputs feed into audience estimates.

What the rules say about the data behind the model

Where a supplier buys audience segments from a broker, the ICO's guidance on organisations using marketing services of data brokers explains what that supplier should be able to demonstrate. Where personal data is processed, the UK GDPR and Data Protection Act 2018 apply, with the Information Commissioner's Office (ICO) as the data protection regulator. Before approval, ask who determines the purposes of processing and request the data source, lawful basis, privacy information and relevant supplier contracts.

For campaigns that also use EU-facing online platforms, the EU Digital Services Act (Regulation (EU) 2022/2065) imposes advertising transparency duties on covered online platforms. It is not an England OOH advertising rule and does not automatically govern roadside or other screen inventory; distinguish the platform's obligations from the rules for the OOH ad itself.

Where AI changes creative and scheduling

Creative variant testing is the most visible application. A digital campaign can rotate several executions and let a model allocate screen time by estimated attention. The risk is that the model optimises for the metric it can see, not the one the brand cares about.

Scheduling carries a similar tension. A model may favour sites with the best predicted footfall, crowding out the contextual placements that made the campaign work.

Comparing AI-assisted and manual planning

AI vs manual planning

AI-assisted

Audience estimate: Footfall + transport
Creative rotation: Dynamic by attention
Post-campaign reach: Modelled incremental
Budget shift: Automated mid-flight
What to check: Source and refresh

Manual

Audience estimate: Panel or historic
Creative rotation: Fixed per site
Post-campaign reach: Recall or lift
Budget shift: Weekly human review
What to check: Confidence and method

AI-assisted approach

Audience estimate: Modelled from footfall and transport data
Creative rotation: Dynamic allocation by predicted attention
Post-campaign reach: Modelled incremental reach
Budget shift: Automated reallocation mid-flight

Manual approach

Audience estimate: Panel or historic site data
Creative rotation: Fixed copy per site
Post-campaign reach: Recall or sales lift study
Budget shift: Weekly human review

What to check

Audience estimate: Source and refresh date
Creative rotation: Metric driving rotation
Post-campaign reach: Confidence interval and method
Budget shift: Who approves the change

What this means for England buyers

For England buyers, supplier due diligence now includes model questions. Ask who owns the training data, how often the model is retrained, and what happens when it is wrong. The out of home advertising market outlook in England sets out how those questions fit a wider buying cycle.

Common questions

Does AI change the advertising rules for out of home in England?

No. The ASA administers the CAP Code for non-broadcast ads, including ordinary OOH, and the BCAP Code for broadcast ads. An AI-assisted OOH plan still has to comply with the CAP Code and, where personal data is processed, the UK GDPR and Data Protection Act 2018.

Can AI replace audience research for out of home?

It supplements rather than replaces it. Modelled audiences are estimates with assumptions attached, so primary research remains the way to test them.

What should a buyer ask a supplier using AI?

Ask for the data source, the model's purpose, the refresh frequency and the human review step. If the supplier cannot name them, treat the audience number as illustrative.

Is AI planning cheaper?

Not necessarily. Licensing audience data and running models carries its own cost, which usually shows up in the rate card.