Local advertising has an awkward economics problem. The media is getting more measurable as the audience becomes harder to find, while the person buying it still has a sensible local budget and a business to run.
A national advertiser can absorb complexity with people, agencies and specialist contracts. A local car dealer, law firm or restaurant cannot. If moving a campaign from linear television into streaming requires four platforms, a data contract and somebody repairing a spreadsheet, the market has failed the buyer.
The money is moving anyway. The IAB estimated that digital video would take 58% of US television and video advertising spend in 2025, up from 29% in 2020. It also linked CTV growth to self-service tools, accessible pricing and better connections between spend and outcomes for smaller businesses. IAB, April 2025 ↗
Local television is not disappearing. Its audience and its commercial model are spreading across broadcast, station apps and other streaming inventory. Nielsen has shown that local streaming can add reach to a linear schedule and that stations are increasingly selling owned and off-property inventory together. Nielsen, November 2025 ↗
Local has inherited a national problem
Fragmentation looks attractive from a distance. More inventory, more targeting and more ways to prove a result. Get closer and somebody has to turn those options into a proposal, price the plan, build creative, activate each line, watch delivery, fix problems and explain the outcome to the advertiser.
Now repeat that for thousands of campaigns, each with a different geography, budget and definition of success. The unit economics get ugly very quickly. A campaign can perform perfectly well and still be a bad business for the media company if it took six people to make it happen.
Madhive is a useful public example of how the market is responding. It describes its platform as a unified system spanning planning, activation, measurement and integration, and says it powers more than 50,000 active campaigns each day. Those are Madhive's figures, not an independent audit, but the operating requirement is clear enough. Madhive ↗
Its announced partnership with Gray Media is also interesting. The stated model combines Gray's local audience relationships with one workflow for planning, activation, optimisation and measurement, plus access to external streaming inventory and audience data. Again, it is a partnership announcement rather than proof of incremental revenue or advertiser outcomes. Gray Media and Madhive, June 2026 ↗
One system is only the start
Replacing a frankenstack with a single operating system removes hand-offs. It does not decide what should be bought. A planning screen can still make an operator choose between a dozen audiences and hundreds of media products without telling them which choice is likely to help the client sell another car.
The real opportunity is a decisioning layer that starts with the advertiser's business and works backwards into media.
The brief might be simple: increase weekday bookings within eight miles of a restaurant, avoid existing customers and do not spend more than $6,000. The system should be able to find suitable audiences and local inventory, estimate trade-offs, assemble the campaign and explain why it made those choices. A person can approve the plan before money moves.
Once the campaign is live, the same layer should compare delivery with the result the advertiser actually cares about. Calls, visits and sales are harder than impressions. They are also the reason the advertiser might renew.
Protocols can remove some of the plumbing
AdCP and AAMP are relevant because a decisioning layer is not much use if every platform requires a bespoke connection. AdCP publishes common tasks and schemas for inventory discovery, media buying, creative, audience activation, accounts and reporting across agent-enabled platforms. It also provides public reference code and versioned specifications. AdCP specification and repositories ↗
AAMP is IAB Tech Lab's broader programme. It combines management protocols with existing advertising standards, reference buyer and seller agents, a registry and the Agentic Real Time Framework. IAB Tech Lab has demonstrated a buyer agent turning a brief into a media plan, negotiating with a seller agent and pushing a confirmed transaction into Google Ad Manager. IAB Tech Lab AAMP ↗
A local media company could use this infrastructure to expose its own inventory, compare external supply and activate an agreed plan without maintaining another collection of brittle APIs. Specialist data and measurement providers could participate without forcing the operator into another interface.
Both efforts have public specifications or reference code that developers can work with now. Neither gives a local operator universal platform coverage, clean customer data or the commercial right to move a client's budget. The rest of the proposition is still a bet.
Local advertisers are not asking for either protocol. Most will never know the names and should not have to. Protocol adoption is a supply-side decision. Customer demand is for an advertising result at a price the business can justify.
The difficult bits sit outside the protocol
First, the data has to be usable at local scale. Inventory needs consistent descriptions, geography and real availability. Audience data needs sufficient coverage in small markets. Outcome data has to connect an ad to a visit, booking or sale without pretending that every match is deterministic.
Then there is the workflow nobody enjoys discussing. Prices change. Creative arrives late. A campaign under-delivers. The client changes the postcode after approval. Any useful agent needs access to those facts, not a polished product catalogue prepared for a demonstration.
Commercial authority is harder again. Someone must decide which inventory the system may buy, how margin is protected, who carries credit risk and what happens when delivery fails. A protocol can transmit the order. It cannot invent permission to place it.
Finally, the governance has to survive contact with real money. Every recommendation and budget change needs an audit trail. Advertiser data must stay within the rights granted for it. Humans need clear approval limits, especially in regulated categories and political advertising. If nobody can explain who was responsible for a decision, the system is not ready to make it.
Start with one market and one outcome
I would not begin by promising an autonomous local advertising platform. Pick one repeatable advertiser category in one market. Connect the media products that are genuinely available, use an outcome that can be observed and let the agent assemble recommendations while a person retains approval.
The early measures are fairly ordinary: time from brief to proposal, campaigns managed per operator, delivery errors, advertiser outcome and retained margin. If those improve, the system can earn more authority. If they do not, adding another agent will not rescue it.
Local advertising improves when a local business can spend a viable amount of money, understand what happened and come back because the advertising worked. The protocols are plumbing.