Most advertisers use AI for optimization but not creative, TripleLift finds

TripleLift this month revealed a world analysis report on how promoting professionals are deploying AI throughout marketing campaign workflows, discovering a pointy cut up between duties the place the expertise has received over practitioners and people the place human involvement stays primarily non-negotiable. The report, titled “The Evolution of AI in International Promoting,” attracts on a survey of 200 promoting trade respondents performed in April 2026 throughout six nations. It was launched on Might 19, 2026.

The findings seize an trade caught between urge for food and apprehension. Most firms now run AI instruments in some a part of their marketing campaign operations, but totally autonomous execution stays the exception reasonably than the norm – and the nearer a activity will get to the inventive act, the extra practitioners pull again.

Broad adoption, shallow roots

In response to TripleLift, 62% of respondents report that their firm remains to be within the exploration or pilot part of AI integration. That determine sits alongside a separate discovering that 60% of firms say they’ve a centralised AI technique. The hole between having a technique and having confidence in it’s vital: fewer than 30% of respondents categorical excessive confidence in that technique. Solely 6.5% describe themselves as “extraordinarily assured,” whereas 49.5% fall within the “considerably assured” band.

These numbers recommend that AI has moved from curiosity to operational fixture at many organisations, however that the transition is shallow. Groups are operating pilots and early deployments, not but institutionalised workflows with clear governance and efficiency benchmarks. The pattern of 200 respondents answered as much as 30 questions every, spanning patrons, planners, and technologists throughout the US (113 respondents), the UK (34), France (31), Canada (17), Germany (3), and Australia (2).

The 4 execution pillars – and the hole

TripleLift frames marketing campaign execution round 4 distinct pillars: media, measurement, viewers, and inventive. Every requires a distinct kind of AI system. In response to the report, the three most prevalent AI pathway combos in use in the present day hyperlink media and measurement collectively, or media, measurement, and viewers in tandem, or measurement and viewers with out media. In all three instances, the inventive pillar is absent.

The report describes inventive as “the absent hyperlink,” noting that it calls for the best stage of human intervention of any of the 4 pillars. The place the opposite three could be linked by way of built-in AI methods to permit near-continuous marketing campaign setup – requiring human approval reasonably than end-to-end guide crafting of offers or belongings – inventive stays a bottleneck that AI has not crossed.

The information tells a selected story

The adoption numbers throughout these 4 pillars are hanging of their unevenness. In response to TripleLift, 73% of respondents at the moment use AI for marketing campaign optimization, protecting bidding, viewers optimisation, and artistic asset refinement. That’s the highest adoption price of any activity class within the survey. Viewers concentrating on comes second at 59%, with AI getting used for viewers segmentation and the development of contextual and behavioural concentrating on segments.

Inventive manufacturing, in contrast, sits at 25%. Even inside that 25%, the sample just isn’t wholesale automation: over half of these utilizing AI within the inventive house deploy it to scale inventive assessments and alter current belongings, whereas fewer than half are producing totally AI-produced outputs. The implication is that AI in inventive is primarily an assistant to human manufacturing reasonably than a substitute for it.

The hole between optimisation at 73% and artistic manufacturing at 25% is likely one of the most concrete knowledge factors within the report. It maps instantly onto the place advertisers draw the road between trusting a machine and insisting on human judgment.

Full automation of marketing campaign execution is rarer nonetheless. Solely 19% of respondents report that AI totally automates their marketing campaign execution course of. On the different finish, 40% nonetheless keep guide management throughout your complete execution workflow. Someplace between these poles, 38% say AI totally automates viewers concentrating on particularly, whereas 31% point out that finances allocation is dealt with totally by AI. Inventive and supply-side levers, in distinction, nonetheless require direct human involvement in most organisations.

The assessment tax

One of many extra concrete and quantifiable findings within the TripleLift report considerations what the corporate calls the “assessment tax” – the time promoting professionals spend checking AI-generated outputs earlier than they go reside. The time saved by AI on execution doesn’t merely return to practitioners as free hours. In response to the report, 45% of respondents spend between one and 4 hours per week reviewing AI outputs. An additional 22% spend between 4 and eight hours on the identical exercise. On the excessive finish of the dimensions, 6% dedicate greater than 12 hours per week to making sure that marketing campaign supplies meet model requirements.

That ultimate determine is price pausing on. Twelve hours per week reviewing AI outputs is, in sensible phrases, a major share of a working week being allotted particularly to catching and correcting what the machine produces. Even for almost all group spending as much as 4 hours weekly, the assessment course of represents a non-trivial dedication – and one which compounds throughout groups and campaigns.

This dynamic just isn’t distinctive to TripleLift’s findings. PPC Land has documented growing advertiser unease with AI-generated content in programmatic environments, with 82% of programmatic patrons now citing AI-powered optimisation as important when choosing companions however concurrently flagging high quality management as a persistent concern. The stress between functionality and oversight is a recurring theme throughout the trade in early 2026.

Why belief is the limiting issue

When TripleLift requested respondents who keep human intervention of their AI-driven workflows to elucidate why, the solutions cluster round a selected set of considerations. Among the many 190 respondents who answered that query (query 28 within the survey), 67% cited lack of belief in AI output as the first motive. Technical errors and hallucinations got here second at 61%. The necessity for model security checks registered at 41%, and knowledge privateness considerations at 21%.

The 67% determine on belief in output is notable as a result of it precedes the opposite considerations in frequency. Model security and technical errors are particular, addressable issues – the type an improved mannequin or higher guardrails can cut back. A normal lack of belief within the output is tougher to resolve. It displays a extra elementary scepticism concerning the reliability and judgment of AI methods in contexts the place a mistake – the mistaken picture, an off-brand phrase, a contextually inappropriate placement – may cause reputational harm earlier than anybody notices.

A January 2026 analysis on PPC Land recognized exactly this sample, documenting that advertisers face real operational uncertainty round AI inventive high quality reasonably than merely technical friction. The TripleLift knowledge supplies a numerical anchor for that remark.

The place AI is already delivering

The report doesn’t current a uniformly cautious image. Within the areas the place AI has discovered footing – primarily data-driven features reasonably than expressive ones – practitioners are seeing measurable enhancements. In response to TripleLift, early adopters report positive aspects throughout three classes: pace (protecting each marketing campaign launch time and optimisation cycle length), effectivity (measured by way of eCPM and cost-per-click), and media metrics together with click-through price and conversion price.

These are the areas the place AI’s numerical and pattern-matching strengths map most instantly onto what promoting operations require. Bidding, for instance, includes processing giant volumes of public sale knowledge sooner than any human crew can handle and adjusting in actual time. Viewers segmentation requires correlating first-party and contextual alerts throughout giant datasets to establish probably responsive teams. Each are duties the place algorithmic methods have demonstrable benefits.

What stays extra contested is whether or not these effectivity positive aspects within the backend truly translate into higher outcomes for manufacturers whose worth is determined by how they convey creatively. The TripleLift knowledge suggests the trade’s working reply, for now, is that they don’t – a minimum of not sufficient to relinquish human inventive management.

The programmatic trade context

The timing of the TripleLift report lands in a interval of concentrated exercise round agentic AI throughout the programmatic provide chain. TripleLift itself introduced its TL Spark agentic intelligence layer in April 2026, positioning it as an orchestration system linking creative, supply, audience, and measurement reasonably than automating any single operate independently. The structure described in that announcement displays the identical four-pillar framework the analysis report maps out.

Elsewhere within the provide chain, PubMatic launched AgenticOS in January 2026 as infrastructure for autonomous marketing campaign execution. The IAB forecasted 9.5% US advert spend progress for 2026, with agentic AI systems cited as a central driver. Mediaocean’s 2026 H1 Outlook Report, launched in January 2026, discovered that entrepreneurs plan to extend AI media spending at a better price than conventional search promoting – a primary within the survey sequence – whereas concurrently reporting problem implementing generative AI instruments internally.

That final discovering mirrors what TripleLift’s survey paperwork on the marketing campaign execution stage. Organisations are allocating budgets towards AI-powered media channels whereas struggling to systematise AI use inside their very own workflows. The exterior AI setting is transferring sooner than the interior one.

The Criteo and TripleLift partnership, reported by PPC Land in April 2026, illustrates how platform-side AI integration is advancing no matter advertiser-side hesitancy. That collaboration combines Criteo’s Commerce Audiences with TripleLift’s curation layer into activatable Deal IDs throughout internet, cellular, and CTV – automating a portion of the availability chain that was beforehand guide. The advertiser should still be reviewing outputs manually, however the infrastructure beneath them is more and more automated.

Survey methodology

TripleLift performed the survey in April 2026. The 200 respondents answered a most of 30 questions every. Geographic distribution was weighted towards the US (113 respondents, or 56.5% of the whole), with the UK second at 34, France at 31, Canada at 17, Germany at 3, and Australia at 2. The report was produced with contributions from Alana Schept (Affiliate Director, International Advertising and marketing), Andrea Strum (Analysis Director), and Emma Sainsbury (Analysis Analyst), amongst others.

The pattern measurement of 200 is comparatively small for a world trade survey, and the geographic skew towards US respondents is price noting when studying any cross-country inferences. The outcomes are indicative of attitudes inside TripleLift’s community {and professional} ecosystem reasonably than a statistically consultant cross-section of your complete promoting trade.

Future priorities

In response to TripleLift, the trade’s wishlist for future AI use centres on 4 areas in ranked order: stock discovery, deal and bundle curation, inventive testing at scale, and AI diversifications to current creatives. The ordering is telling. The very best-priority gadgets are supply-side and transactional – features closest to the place automation already works. Inventive stays final, even within the aspirational view. “AI diversifications to current creatives” is a extra restricted ambition than totally AI-generated manufacturing, suggesting the trade envisions a future the place AI adjusts and scales human-made inventive reasonably than originating it.

Timeline

Abstract

Who: TripleLift, a supply-side platform and Inventive SSP, surveyed 200 promoting professionals throughout the US, United Kingdom, France, Canada, Germany, and Australia as a part of its 2026 AI analysis report.

What: The report “The Evolution of AI in International Promoting” paperwork a large adoption hole between AI use in marketing campaign optimisation (73% of respondents) and AI use in inventive manufacturing (25%), with solely 19% of respondents totally automating marketing campaign execution. It identifies a “assessment tax” the place 45% of practitioners spend as much as 4 hours per week reviewing AI-generated outputs, and finds that 67% cite lack of belief in AI output as the first motive for sustaining guide oversight.

When: The survey was performed in April 2026. The report was revealed on Might 19, 2026.

The place: The survey covers promoting professionals throughout six nations: the US (113 respondents), United Kingdom (34), France (31), Canada (17), Germany (3), and Australia (2). The findings apply throughout programmatic promoting contexts together with open internet, cellular, CTV, and retail media.

Why: The report issues for the advertising and marketing neighborhood as a result of it quantifies the hole between the place AI functionality has reached and the place practitioner belief lags behind. The inventive pillar – the one most linked to model identification and client notion – stays the least automated element of marketing campaign execution, at the same time as data-driven features have reached excessive adoption. For practitioners navigating AI device adoption choices, finances allocation between human and automatic manufacturing, and inner governance round AI workflows, the numbers present a concrete reference level grounded in trade peer behaviour.