Brands Using AI for Advertising 2026

TL;DR — Key takeaways

When a brand announces it’s “using AI for advertising,” it usually means Level 1 (automated bidding) — not the generative creative or autonomous campaigns the press release implies.
The AI Advertising Maturity Model has four levels: automated bidding (90% of brands), audience intelligence (40%), AI-assisted creative (15%), generative creative (under 5%).
JPMorgan Chase’s contract with Persado for AI copywriting — where machine-generated copy outperformed human writers by over 450% in click-through rates — remains the most documented enterprise case study in this space.
Heinz’s “Draw Ketchup” campaign (2022) proved that brand memory can be so strong that AI image generators reproduce it unprompted — a smarter insight than any programmatic optimization.
The real opportunity most brands miss isn’t generative AI for creative — it’s using AI to unify audience intelligence across channels before spending a dollar on creative production.

Every major brand communications now includes a sentence about AI. Press releases announce AI-powered campaigns. Award entries celebrate AI-driven personalization. Marketing conferences dedicate entire tracks to brands using AI in advertising. Almost none of it distinguishes between a Google Smart Bidding algorithm making bid adjustments every 100 milliseconds and a brand genuinely using machine learning to generate creative, predict customer lifetime value, or orchestrate cross-channel campaigns autonomously.

That distinction matters enormously — in budget, in sophistication, in organizational readiness, and in actual results. What follows is an honest categorization of how brands actually use AI in advertising today, with case studies that survived scrutiny rather than press release mythology.

Table of Contents

Toggle

The AI Advertising Maturity Model: four levels that most coverage conflates

Level 1 is automated bidding. This is what 90%+ of brands mean when they say they’re using AI for advertising. Google’s Smart Bidding, Meta’s Advantage+ campaigns, Amazon’s dynamic bids — these systems adjust bids in real time based on predicted conversion probability, using ML models trained on aggregate platform data. They work. They often outperform manual bidding at scale. But they are platform-operated algorithms, not brand AI strategy. The brand’s role is setting targets and constraints, not building the intelligence.

Level 2 is audience intelligence. Roughly 40% of sophisticated advertisers operate here — using first-party data, lookalike modeling, predictive lifetime value scoring, and behavioral segmentation to build more accurate audience targets than platform defaults offer. This is where brands that have invested in clean CDPs, data clean rooms, and ML modeling genuinely differentiate. The output is better audience targeting, not different creative.

Level 3 is AI-assisted creative. Approximately 15% of brands — mostly enterprise with dedicated MarTech stacks — use AI tools to test copy variations, optimize dynamic creative elements, predict which headline/image combinations will outperform, or personalize landing page copy based on the ad that drove the click. This is where Persado operates, where Dynamic Creative Optimization platforms like Adobe Target and Optimizely intersect with advertising, and where AI starts visibly touching what the customer actually sees.

Level 4 is generative creative. Under 5% of brands, mostly large enterprise or bold experimenters. AI systems that generate ad copy, images, or video at scale — not to replace human creative direction, but to produce more variations faster for testing. Coca-Cola’s “Create Real Magic” platform, various brands experimenting with Midjourney and DALL-E for concept development, and early movers using Runway for video ad production all sit here.

450%

higher click-through rate achieved by Persado AI copy vs. human-written copy

JPMorgan Chase enterprise case study — the most documented AI copywriting result in financial services

Case studies that actually hold up: what brands are doing and what it proves

Epinium data

Based on campaigns we’ve managed across 12+ European Amazon marketplaces, brands that implement AI bid optimization see ACoS improvements of 18–35% in the first 60 days.

JPMorgan Chase + Persado (Level 3). JPMorgan Chase signed an enterprise contract with Persado in 2019 — one of the earliest and most-documented commitments to AI copywriting at enterprise scale. Persado’s system generates copy variations using a language model trained specifically on emotional language and its relationship to conversion outcomes. In Chase’s reported results, machine-generated copy outperformed human-written control copy by over 450% in click-through rates in some tested variants. The finding that generated the most internal discussion wasn’t the performance uplift — it was that the AI consistently chose language patterns that human copywriters consistently rejected as “too simple” or “not our tone.” This is the uncomfortable signal: AI sometimes wins not despite ignoring brand instincts but because of it.

Heinz “Draw Ketchup” (Level 4, but the insight is deeper). In 2022, Heinz ran a campaign asking AI image generators to produce images of “ketchup.” Every result looked like Heinz — red bottle, distinctive cap, characteristic label proportions. The campaign was smart not because of the technology but because of the strategic insight it revealed: Heinz’s brand memory is so deeply embedded in training data (and by extension, in collective human visual memory) that an AI with no brand brief spontaneously reproduces it. That’s a different kind of AI advertising insight than campaign automation — it’s using AI as a brand health diagnostic. The executional campaign that followed was modest; the strategic implication was significant.

Coca-Cola “Create Real Magic” (Level 4). In 2023, Coca-Cola opened a platform where artists and fans could create original content using GPT-4 and DALL-E, with access to Coca-Cola’s brand archive as source material. The campaign positioned AI as a collaborator with a creative community rather than a replacement for creative production. It generated substantial earned media and positioned Coca-Cola as an AI-forward brand without requiring the company to actually automate its core advertising production. Strategically, it was less about AI-generated ads than about AI-adjacent brand positioning at a moment when the category was getting maximum media attention.

Spotify Personalized Ads (Level 2). Spotify’s advertising platform uses listening behavior data — genres, artists, playlist context, listening time patterns — to build audience segments that advertisers can target with unusual precision. The insight here isn’t the ad tech; it’s the signal quality. A brand targeting “people who listened to workout playlists on weekday mornings in the last 30 days” has behavioral context that demographic targeting cannot replicate. This is Level 2 AI in action: the creative is human, but the audience intelligence is machine-generated from behavioral data at scale. Spotify’s own research shows that contextually targeted podcast ads drive 24% higher brand recall than non-targeted placements.

FREE SESSION

Which AI advertising level is right for your brand?

We map your current data infrastructure, creative capacity, and campaign goals to identify which maturity level creates the highest-ROI next step — and what it actually takes to get there.

Book session → ✓ Free ✓ 30 min ✓ No pitch

What most brands get wrong about AI advertising strategy

The most common mistake isn’t choosing the wrong AI tool — it’s skipping Level 2 to chase Level 4. Brands that invest in generative AI creative before they have a functioning first-party data strategy are spending on the visible, sexy layer while leaving the most valuable layer untouched.

Here’s why this matters. The ROI ceiling of AI-generated creative is determined by how precisely you can target it. A beautiful AI-generated ad shown to the wrong audience is still a wasted impression. Generative AI multiplies creative variation; it does not fix targeting. Audience intelligence — the Level 2 work of building predictive LTV models, behavioral cohorts, and cross-channel identity resolution — is what makes every downstream AI investment, including creative, more efficient.

What we see at Epinium with brand advertisers is a consistent pattern: the brands achieving the highest measurable ROI from AI advertising are not the ones with the most sophisticated creative AI. They’re the ones who cleaned their first-party data, built proper audience segments from behavioral signals, and then applied even Level 1 automated bidding with dramatically better input signals. Clean data feeding good algorithms beats expensive AI creative feeding mediocre targeting.

A McKinsey analysis on AI in marketing found that companies using AI-driven personalization at scale report 10-15% revenue uplift — but the key variable wasn’t the sophistication of the AI; it was the quality and completeness of the customer data the AI was trained on.

AI advertising maturity: comparison by level

Level	What it is	Brand adoption	Key requirement
1 — Automated bidding	Platform ML (Smart Bidding, Advantage+)	90%+	Conversion tracking, sufficient data volume
2 — Audience intelligence	First-party data, LTV prediction, behavioral cohorts	~40%	Clean CDP, data science capacity
3 — AI-assisted creative	Copy testing, DCO, Persado-type optimization	~15%	Volume of creative variants, MarTech stack
4 — Generative creative	AI-generated copy, images, video at scale	<5%	Brand safety controls, legal clearance, creative team

Frequently asked questions

Which brands are using AI for advertising most effectively?

The most documented enterprise cases are JPMorgan Chase (AI copywriting via Persado), Coca-Cola (generative creative experimentation), Heinz (AI as brand intelligence tool), and Spotify (behavioral audience intelligence). Effectiveness depends heavily on the maturity level: brands achieving consistent measurable ROI tend to be those who have mastered Level 2 audience intelligence before investing in Level 3 or 4 creative AI. The mistake is treating generative creative as the entry point when audience intelligence is the higher-ROI investment for most brands.

How is AI used in digital advertising specifically?

Four main applications: automated bidding (platform algorithms adjusting bids in real time based on predicted conversion probability), audience segmentation and lookalike modeling (ML identifying high-value audience clusters from first-party data), dynamic creative optimization (AI selecting which creative elements — headline, image, CTA — perform best for each user segment), and generative content creation (AI generating copy or visual concepts for testing at scale). Most brands operate at level 1 or 2; levels 3 and 4 require substantially more data infrastructure and organizational readiness.

What is Persado and how do brands use it for advertising?

Persado is an AI platform that generates marketing copy by analyzing the emotional and functional language patterns that correlate with conversion outcomes. Instead of generating language like a general-purpose LLM, Persado’s system is trained specifically on the relationship between message characteristics and purchase behavior. Brands like JPMorgan Chase use it to generate and test copy variants at a scale impossible with human copywriters. The JPMorgan case is the most public: machine-generated copy reportedly outperformed human control copy by over 450% in click-through rates on some tested variants — the disorienting part being that human reviewers often rated the winning AI copy as less compelling before seeing the results data.

Is AI-generated advertising creative actually effective?

The evidence is mixed and context-dependent. For direct response advertising where click-through rate and conversion rate are clear metrics, AI-optimized copy (Level 3) has strong documented evidence of outperforming human copy in controlled tests. For brand advertising — awareness, recall, emotional resonance — generative AI creative (Level 4) has less evidence of consistent outperformance, and several high-profile AI creative campaigns have underperformed on brand metrics despite novelty. The pattern that emerges: AI wins on direct response optimization; human creativity still leads on brand building.

How should brands start using AI for advertising in 2026?

Start at Level 2, not Level 4. Before investing in AI creative tools, assess the quality of your first-party data: do you have clean customer purchase history, behavioral signals beyond transactions, and some form of identity resolution across channels? If not, that is your highest-ROI AI investment. Level 1 (automated bidding) is already operating in almost every campaign — optimize it by improving the signal quality going in. Level 2 audience intelligence work — building predictive LTV models, suppression lists, behavioral cohorts — then multiplies the impact of every Level 3 and Level 4 investment that follows.

The brand that generates the most press about AI advertising is not necessarily the brand getting the most value from it. Generative creative campaigns win awards and generate earned media. First-party data infrastructure improvements rarely make it into trade press. But ask any performance marketer with a genuine view into both: the data work is where the money is made.

TRANSFORM BY EPINIUM

Build the AI advertising foundation that actually moves revenue

Brands that work with Epinium move from Level 1 to Level 2-3 AI advertising maturity in 90 days — with measurable improvement in ROAS and new-to-brand customer acquisition cost.

Book free session →

Free · 30 min · No commitment

What Actually Changed in 2025-2026

Amazon Rufus scale (Q4 2025)

Amazon Rufus reached 300M active users and drove roughly $12B in incremental annualized sales per Amazon Q4 2025 earnings — shifting discovery from keywords to conversational intent.

Buy for Me launch (April 2025)

Amazon’s Buy for Me feature lets Rufus purchase from external sites on the user’s behalf, normalizing agentic commerce outside walled gardens.

Checkout embedded in ChatGPT (late 2025)

OpenAI shipped in-chat checkout with partner merchants, forcing brands to treat ChatGPT as a distribution channel, not only a research tool.

Google AI Overviews + E-E-A-T tightening (2025)

Google’s 2025 core updates penalized low-differentiation AI content and rewarded first-party experience signals — raising the bar for editorial AI workflows.

What percentage of “AI-driven” ad claims hold up under audit?

In our audits of 300+ accounts, roughly 30% of “AI-driven” claims match actual model-driven decisions; the other 70% are rule-based automation rebranded. Ask vendors for the feature flag log before signing.

When is AI advertising NOT worth it for a brand?

Monthly ad spend under $30K, single-product catalog, or no first-party data — AI optimizers cannot outperform broad match + manual negatives at that scale. Fix targeting fundamentals first.

How should a brand measure AI ad uplift honestly?

Geo-split or holdout tests over 6+ weeks. Platform-reported “AI lift” numbers are reconciled against the baseline the platform itself chose — always flattering. Geo-holdouts give an unbiased read.