Where does AI fit in programmatic media planning, and where does a planner stay in charge?
Part of Programmatic media planning
The short answer
AI media planning uses software models to handle the data heavy steps of a programmatic plan: gathering candidates, segmenting, forecasting and drafting. IAB and PwC describe the shift as early, with progress mostly in agency operations rather than fully autonomous buying. Planners keep the brief, the final site list, client claims and spend approvals.
Key takeaways
- IAB and PwC's April 2026 report says agentic media buying efforts are in their early days, with progress concentrated in agency operations and optimization.
- In IAB's 2025 State of Data survey of 529 companies, 70 percent had not yet fully scaled AI across planning, activation and analysis.
- The same survey found respondents rated AI weakest where human judgment and strategy are required, and strongest at segmentation and data aggregation.
- An AI assisted workflow needs a written brief of its own: inputs, task, output format, acceptance test and the person who signs off.
- Accountability stays with the agency and the advertiser, not with the tool that drafted the recommendation.
How is AI used in programmatic media planning today?
AI in a media plan is a set of tasks that used to be manual and are now partly automated: pulling performance history, grouping audiences, estimating delivery, drafting proposals and suggesting placements. The IAB State of Data 2025 report (opens in a new tab), published in March 2025 from a survey of 529 agencies, brands and publishers, maps these tasks across three stages of the campaign lifecycle: planning, activation and analysis.
Definition
- AI assisted media planning
- A planning workflow in which software models draft or score parts of the plan, such as candidate sites, forecasts or budget splits, and a named person reviews, edits and approves the result before anything is bought.
| Planning task in the IAB list | What the model contributes | What the planner still checks |
|---|---|---|
| Identify and segment audiences | Groups people or contexts from first party and partner data | Whether the segments match the brief and use only data the client may use |
| Determine media mix, pacing and pricing | Simulates budget splits across channels and placements | Whether the scenarios respect the client's fixed commitments and flight dates |
| Forecast campaign inventory and outcomes | Estimates reach and delivery from historical patterns | Which history the forecast drew on and whether it resembles this campaign |
| Recommend ad placements and formats | Ranks candidate sites, channels or formats | The evidence behind each placement and anything left off the list |
| Select partners and evaluate RFPs | Scores proposals against stated criteria | Whether the criteria are the client's, not the tool's defaults |
| Automated gathering of market intelligence | Summarizes trends and competitor activity | That each cited fact has a source someone opened |
Forecasting is the clearest case of a model that planners already rely on. Google's Display & Video 360 planning help (opens in a new tab) says its forecasts "use trends in the ad market, including the historical performance of similar campaigns to model performance," drawing on a recent period equal to the campaign's date range for up to 92 days. A forecast built that way is only as relevant as the campaigns it resembles, which is why the planner, not the model, decides whether this brief is like those.
What does the IAB and PwC report say about AI in planning?
The IAB and PwC Internet Advertising Revenue Report for full year 2025 (opens in a new tab), released in April 2026, names AI as the first of five trends for the industry. Its introduction says generative AI is reducing creative production costs, "while agentic systems are also changing how companies buy, plan, optimize, and execute media." The trends section then sets a limit on how far that change has gone.
These efforts, however, are in their early days. Most progress is occurring in agency-side operations and optimization rather than fully autonomous buying, as the industry continues to evaluate both the potential and risks of greater automation.
The IAB's earlier survey gives the baseline behind that caution. In the State of Data 2025 report, 70 percent of agencies, brands and publishers had not yet fully scaled AI across media planning, activation and analysis. Respondents rated AI highest for audience segmentation and data aggregation, and lowest for ad fraud prevention, brand safety and contract management. The report's own heading for that finding reads that AI "falls short today where human judgment and strategy are required." Treat these as 2025 survey results from one respondent group, not a measure of the whole market in 2026.
Which planning steps should keep human review?
A useful test is reversibility and reach. A step whose mistake is cheap to catch later, such as a first pass of candidate sites, can run with light review. A step whose mistake reaches the client, the budget or the public needs a person to sign off. The State of Data 2025 recommendations say the same in general terms: "Require human oversight to verify AI inputs and outputs, mitigate bias, and ensure compliance with evolving privacy regulations."
| Step | Review level | Why |
|---|---|---|
| Interpreting the client brief | Sign off required | A misread market, language or avoid topic changes the steps that follow |
| Building the candidate pool | Spot check a sample | Errors here are caught again at site review |
| Writing the reason for each site | Review each item | The reason is what the client reads and challenges |
| Forecasting delivery | Review each item | The planner decides whether the historical campaigns resemble this one |
| Budget split and flight dates | Sign off required | It commits money and must match the insertion order |
| Formatting the export file | Automate with logs | A file check against the platform format catches mistakes |
How should an agency brief an AI assisted workflow?
An AI step needs its own brief, separate from the client brief. Without one, the tool fills gaps with defaults nobody chose. The IAB's State of Data 2025 Companion Guide (opens in a new tab) says AI can support "media partner selection and RFP evaluations by predicting success based on historical data, brand alignment, and projected ROI." Each of those inputs is a choice: whose history, which definition of brand alignment, whose projection. The workflow brief records those choices.
Six parts of a brief for an AI assisted planning step
- Name the task in one sentence. For example, rank candidate websites against the confirmed client brief for a Nevada audience.
- List the allowed inputs. Name the confirmed brief, the data sources the step may read and any data it must not touch, such as client customer files.
- Fix the output format. Ask for a ranked list with a reason and supporting pages for each site, plus a separate list of sites left out with reasons.
- Write the acceptance test. State what a reviewer checks before accepting the output, such as opening one supporting page per site and confirming the market and language.
- Name the reviewer and the sign off. One person reviews the output; a second person with approval rights signs off before anything reaches a buying platform.
- Record the version. Save the inputs, the output and the date, so the plan can be explained if the client asks months later.
Before an AI output enters the plan
- The output answers the confirmed brief, not an earlier draft.
- Each recommended site or channel shows evidence a person can open.
- Anything the tool left out is listed with its reason.
- Forecast figures name the period and campaigns they came from.
- The reviewer's name and the review date are recorded.
- No client claim in the proposal rests only on the tool's wording.
What should a planner never delegate to AI?
Some decisions belong to people because someone has to answer for them. Canada's federal, provincial and territorial privacy authorities put it directly in their principles for generative AI (opens in a new tab): "Know that accountability for decisions rests with the organization, and not with any kind of automated system used to support the decision-making process." The same document asks developers to make outputs "traceable and explainable."
- Agreeing the brief with the client. The tool can structure it; only the client relationship owner can confirm it.
- Approving the final list. An approval is a promise to the client that someone looked.
- Committing spend. Budgets, insertion orders and deal terms carry contract obligations.
- Claims to the client or the public. A statement about audience, reach or suitability needs a person who checked its basis.
- Deciding what personal data a campaign may use. That decision sits with the advertiser and its agency under the privacy law that applies to them.
For the related questions of skew in automated recommendations, what makes an explanation real and when to overrule a suggestion, see the guides on AI bias in media planning, explainable AI in advertising and overriding AI recommendations.
How does PlacementGraph use AI in planning?
What does an AI assisted plan look like for one brief?
Fictional example
A science center's winter membership drive in Nevada
An unnamed science center with sites in Las Vegas and Reno asks its agency for a winter display plan to sell family memberships, in English, beside parenting, science news and local family activity coverage. The agency writes a workflow brief before any tool runs: the AI step may read the confirmed client brief and public site evidence, must not touch the center's member list, and must return a ranked list with reasons and a separate list of sites left out.
The planner confirms the structured brief with the account lead, who catches that the tool read "Nevada" as statewide while the client wants the two metro areas only. The candidate pool is spot checked on a sample. Each site reason is reviewed, and two sites whose supporting pages turn out to be event listings rather than articles are removed and logged.
The forecast is reviewed against the period it drew on, and the planner notes that the history did not include a membership campaign of this kind. The approver signs off the list, the budget split and the flight dates. The proposal to the client states which steps were automated and who reviewed each one.
Frequently asked questions
Can a general purpose chatbot build a programmatic site list?
It can suggest names, but a suggestion is not a vetted list. A chatbot answer does not show which pages it read, whether a site sells ad space programmatically or whether it has an audience in the target market. Treat its output as a candidate pool and check each site like any other.
Should an agency tell the client which planning steps used AI?
Saying so answers a concern the IAB survey recorded: about half of brands worried about a lack of transparency into how partners use AI on their behalf. A short note in the proposal listing the automated steps and the reviewer for each keeps the record simple for both sides.
Does AI planning change who is accountable for the plan?
Canada's privacy authorities state that accountability for decisions rests with the organization, not with the automated system that supports the decision. Under that principle, the agency that delivers the plan and the advertiser that approves it remain responsible for what it contains and what it spends.
Can an AI forecast replace the buying platform's own forecast?
Use both for different jobs. A planning model can compare scenarios quickly, while the buying platform's forecast reflects the inventory and settings that platform will actually bid on. When the two disagree, record both and the reason, and plan against the more conservative figure until delivery data arrives.
Sources
- Internet Advertising Revenue Report, Full Year 2025 (opens in a new tab)IAB and PwC, April 2026
- State of Data 2025: The Now, the Near, and the Next Evolution of AI for Media Campaigns (opens in a new tab)IAB, March 2025
- State of Data 2025 Companion Guide (opens in a new tab)IAB, March 2025
- Principles for responsible, trustworthy and privacy-protective generative AI technologies (opens in a new tab)Office of the Privacy Commissioner of Canada
- Create a plan for a campaign (opens in a new tab)Google, Display & Video 360 Help
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.