Digital Marketing
August 31, 2026
Last updated October 20, 2026
Estimated Read Time: 7 minutes
Most marketers can't prove AI is working because they use it in too few places, point it at goals nobody can audit, and can't measure what it does. In AdPipe's Pulse Report 2026, 78% of teams couldn't point to clear, measurable results, even though none were sitting AI out.
The Pulse Report calls this the proof gap: the distance between the value teams feel from AI and the value they can measure. According to the report, the proof gap is the defining AI story for marketing teams in 2026. Every team surveyed had adopted AI, and 71% use it regularly or have it built into daily work. Adoption isn't the problem. Evidence is.
For most marketing teams, AI is hard to measure rather than failing. In AdPipe's Pulse Report 2026, 64% of teams said AI delivers some value that's hard to measure, while only 7% said it has delivered little to none.
The last three rows add up to the 78% who can't point to clear results. Most of that group feels AI helping and can't put a number next to it. That group needs something to count before it needs a new tool.
Time is the most common blocker. In the Pulse Report, 27% of teams named lack of time or resources to implement as their single biggest obstacle, followed by not knowing which tools are worth it (22%) and concerns about brand safety or quality (19%).
The rest of the list:
One respondent put the time problem plainly:
"I feel like the concept of AI interests executives, but for the people doing the work it's still very manual and time consuming to set up and requires a lot of oversight." Marketing individual contributor, Pulse Report 2026
Leadership buy-in is the least-cited blocker at 5%. For most teams, the permission they're waiting for probably already exists. What's missing is the hours to set things up and a clear way to tell which tools earned their place.
Each adoption stage hits a different wall. In the Pulse Report, teams experimenting with AI are slowed most by brand safety concerns, teams using it regularly are slowed by time, and teams with AI built into daily work are slowed by measurement. The share that can prove results rises from 8% to 46% along the way.
Teams at the top of the table are earlier on the same road. The obstacle moves from fear, to time, to measurement, and each one needs a different fix.
Want the full breakdown by stage? Get the Pulse Report.
Teams that can prove AI is working use it in more places and aim it at a metric they're already judged on. In the Pulse Report, 43% of teams running seven or more AI use cases can prove results, compared with none of the teams running one or two.
Depth is the clearest line in the data. At one or two use cases, 0% can prove results. At three to six, 23% can. At seven or more, 43% can. Five tools pointed at the same task still count as one use case.
Aim matters too. Teams that point AI at a metric they're already measured on prove results 40% of the time. Teams whose stated goal is saving time prove the least.
"When used correctly, it's an awesome tool. When used just to save time it creates slop and business none of us need." Creative and design manager, Pulse Report 2026
For a practical walk-through of the first week, see Modular Gave You Ideas. Monday Is Where You Prove Them.
Use the metric your team is already judged on. In the Pulse Report, 59% of marketing leaders said revenue or pipeline contribution matters most this year, ahead of conversion rate and brand awareness at 41% each. Pointing AI at an existing scorecard metric makes the result much easier to defend.
Respondents could choose more than one metric, so the column adds up past 100%. The report's advice is simple: don't invent a new AI metric. Borrow the one already on your scorecard, and the argument gets much shorter.
Creative is where the proof gap gets concrete, because marketers know which ad won but rarely know why. Only 12% of teams in the Pulse Report use AI to understand why their creative performs, the least-adopted use case in the survey, while 83% use AI for copy.
When the Pulse asked what teams most want AI to do reliably, about half the wish list described the same thing: tell me why this creative performed, show me what to change, and connect it to revenue.
That's a diagnosis problem. AdPipe approaches it by reading each ad frame by frame and tying what's inside it to performance data, which gives a team something specific to count from ads it has already run. For more on the difference between reporting results and explaining them, see What Is the Difference Between Creative Analytics and Creative Intelligence? We've also covered the broader shift from AI hype to proof in The First AI Boom Is Over. Marketers Are Hungry For Proof.
Start by measuring one thing well. The Pulse Report recommends picking one recurring decision, capturing a rough baseline, and reporting the result every month, even if the first numbers are imperfect.
Pick a single decision to measure. Not the whole program. One recurring decision, like which creative to run or which variant to scale. Then track what happens after it.
Accept a rough baseline. "Hard to measure" usually means "never baselined." An imperfect before-number beats an elegant nothing.
Report it monthly, even badly. A rough number reported every month beats a perfect number reported never.
Budget the setup hours. The top blocker is time to implement. Treat instrumentation as part of the rollout cost, and name who owns it.
Write the brand guardrail down. A named reviewer and a short written standard unblocks more work than another round of debate.
On the creative side, the quickest start is last quarter. Pull your top five and bottom five ads and list what's inside each one: the hook, the first frame, the pacing, the offer. See what the winners share.
See what's inside your own top and bottom performers. Get your free creative insights
The figures in this article come from AdPipe's Pulse Report 2026, a survey of marketing, creative, and agency leaders completed between July 7 and August 26, 2026. Respondents came from companies ranging from under $50 million to over $5 billion in revenue, and more than three in four answered from manager level or above. Results are unweighted.
Respondents opted in to an AI survey, which likely over-represents engaged teams. The report treats the proof gap as understated rather than overstated, and treats single-digit differences between groups as noise.
Why can't marketers prove AI is working?
Most marketers use AI in too few places, aim it at goals nobody can audit, and can't measure the results. In AdPipe's Pulse Report 2026, 78% of teams couldn't point to clear, measurable results, and 64% said they see value they can't measure.
Does that mean AI isn't working for marketing teams?
No. Only 7% of teams in the Pulse Report said AI has delivered little to none. The largest group, 64%, sees some value but finds it hard to measure. The gap is in proof, not in whether AI helps.
What is the biggest blocker to getting value from AI?
Lack of time or resources to implement, named by 27% of teams in the Pulse Report. Not knowing which tools are worth it came next at 22%, followed by brand safety or quality concerns at 19%. Only 5% named leadership buy-in.
How many AI use cases does it take to prove results?
Depth matters more than any single tool. In the Pulse Report, no team running one or two AI use cases could prove results. At three to six use cases, 23% could. At seven or more, 43% could.
What metric should marketers use to prove AI's value?
Use a metric the team is already judged on. In the Pulse Report, 59% of marketing leaders named revenue or pipeline contribution as the business metric that matters most this year, and teams that aim AI at a metric they're measured on prove results 40% of the time.
Where should a marketing team start proving AI's value?
Pick one recurring decision, such as which creative to run, capture a rough baseline, and report the result monthly. On the creative side, list what's inside last quarter's top and bottom ads to see what the winners share.
October 15, 2026
September 2, 2026
