Writing

• AI in Marketing
• 23 September 2026

What is AI in marketing? Types, examples, benefits and risks

AI in marketing is the use of software that learns from data to do marketing work that used to need a person: predicting who will buy, deciding which ad to show and what to bid, writing and designing content, answering customers, and increasingly running whole campaign steps on its own. In practice it sits inside tools most teams already use, from Google Ads to their CRM.

The money is already moving. Gartner’s 2026 CMO Spend Survey of 401 marketing leaders in North America, the UK and Europe found they now put 15.3% of their marketing budgets into AI, yet only 30% say their AI capabilities are mature enough to scale. So most of us are spending like leaders and operating like beginners, which is a very marketing way to adopt anything.

What is AI in marketing, in plain English?

Strip away the vendor language and AI in marketing is pattern-matching at a scale no team could manage by hand. A model looks at a large amount of past data (clicks, purchases, emails opened, words written by humans) and uses it to make a prediction or produce something new. Everything else is a variation on which data it learned from and what it is allowed to do with the answer.

None of this is new in online marketing. Netflix wrote up how it uses algorithms to choose a different thumbnail for each member back in December 2017, and ad platforms were setting bids by machine long before anyone had heard of ChatGPT. We just used to call it “the algorithm” and blame it when cost per click went up. What changed after 2022 is that generative models made AI something marketers could talk to directly, so it jumped from the media team’s black box to everyone’s browser tab.

What are the main types of AI used in marketing?

It helps to split AI in the marketing industry into four families, because they fail in different ways and need different kinds of supervision.

Predictive AI is the oldest of the four and still does most of the heavy lifting in media, while generative AI is what people usually mean when they say “AI” in a meeting. The agentic kind is the one vendor decks get most excited about, and your legal team least. The difference between the last two matters enough that I’ve written it up separately in generative vs agentic AI in marketing.

How is AI used in marketing? Examples across the marketing mix

The same company will usually use all four types in different jobs, so the application of AI in marketing makes more sense sorted by job.

Insight, analytics and segmentation

AI in marketing analytics mostly means models that find groups and signals humans would miss or take weeks to find: which customers are drifting, which combinations of behaviour predict a second purchase, which campaign actually drove the sale. The practical upgrade from generative tools is that you can now ask a question of a dataset in English and get a first cut back in minutes, which is lovely until it miscounts a column without mentioning it (so you still check the numbers). I’ve gone through how to do this properly, including what data you need, in AI customer segmentation.

Media buying and targeting

This is where most marketing budgets already meet AI, whether the team thinks of it that way or not. Google describes Performance Max as a goal-based campaign that uses Google AI “to make more accurate predictions about which ads, audiences and creative combinations perform best”, with Smart Bidding deciding placements across YouTube, Display, Search, Discover, Gmail and Maps. You supply the budget, the conversion goal and the assets, and the machine decides the rest. The catch is that these systems optimise hard towards whatever conversion you define, so if you tell them a newsletter sign-up is a conversion, you will get a great many newsletter sign-ups.

Personalisation

Netflix’s engineering team explained that it picks which artwork to show each member using contextual bandit algorithms, so someone who watches a lot of romantic films might see Good Will Hunting with Matt Damon and Minnie Driver, while a comedy fan sees Robin Williams. The team described it as the first time Netflix had personalised how it recommends a title as well as which titles it recommends (Netflix Technology Blog), which is a useful way to think about personalisation generally: the offer can stay the same while the way you present it changes per person.

Content and creative production

Unilever is probably the clearest large-scale case. In March 2025 it said it was building 3D “digital twins” of products such as Dove, Vaseline, TRESemmé and Clear, using Nvidia’s Omniverse platform with the creative-tech studio Collective World, so AI can generate product imagery without a fresh shoot. Unilever reported imagery produced twice as fast and 50% cheaper, and a TRESemmé pilot in Thailand with an 87% cut in content creation costs. Its Chief Growth and Marketing Officer, Esi Eggleston Bracey, was careful to add that pushing out more content was never the aim, because anyone can do that, which is a line more of us should have on a Post-it.

Personalised creative at scale predates ChatGPT, too. For Diwali 2021, Mondelez India’s Cadbury, with Ogilvy India, Wavemaker India and the AI video company Rephrase.ai, let local shopkeepers use AI to make their own version of a Shah Rukh Khan ad that promoted their shop by name, covering 2,000+ retailers across 500 pin codes (LBB Online).

AI in marketing and branding

Brand advertising is where AI is most tempting and least proven. Coca-Cola worked with the San Francisco studios Secret Level and Silverside AI to rebuild its 1995 “Holidays Are Coming” spot with generative tools for its 2025 holiday campaign, while stressing that the music was “performed and sung by real artists”. Heinz went the other way with its “A.I. Ketchup” campaign, asking an image generator what ketchup looked like and making the openness of the process the idea itself (Adam Peruta, whose research comes up below, describes it in HBR). Both are useful to study, but neither tells you whether AI brand work actually sells, and a press release was never going to.

Customer service and conversational marketing

Klarna, the Swedish buy-now-pay-later company, is the case everyone quotes, and it’s worth quoting both halves. In February 2024 it said its OpenAI-powered assistant had handled 2.3 million conversations in its first month, two-thirds of its customer service chats, doing the work of 700 full-time agents, cutting time to resolve from 11 minutes to under 2, and forecast a $40 million profit improvement for 2024. By May 2025 the chief executive, Sebastian Siemiatkowski, was telling Bloomberg that “really investing in the quality of the human support is the way of the future for us”, and Klarna was piloting a scheme to recruit human agents again, which I read as a polite way of saying they’d overdone it.

Campaign operations and agents

The agentic end is newer and mostly still in vendor announcements. Salesforce launched Marketing Cloud Next in June 2025 with agents that it says can generate a campaign brief, build the audience, draft email and SMS copy, set up the customer journey, qualify leads around the clock and recommend paid media changes. McKinsey’s State of AI in 2026 survey (1,719 respondents, published August 2026) found that respondents in consumer goods and retail most often report using agents in marketing and sales.

Less glamorous but arguably more valuable is the back-office work. Wayfair, the US online furniture retailer, embedded OpenAI models into its catalogue of roughly 30 million items and corrected 2.5 million product tags, according to Masha Shunko (University of Washington Foster School of Business) and Serguei Netessine (Wharton) in HBR. The authors point out that inconsistent product attributes damage search, recommendations, merchandising and returns, so this is marketing even if nobody puts it in the awards entry.

What are the benefits of AI in marketing?

• Speed and cost of production. Unilever’s twice-as-fast, 50%-cheaper imagery is the kind of number you can plan around, and resizing, versioning and localisation are the least controversial wins going.
• Personalisation at a scale no team could staff. Cadbury could not have shot 2,000 shop-specific ads with a real film star, and Netflix cannot hand-pick artwork for every member.
• Better decisions in media. Setting bids and placements across Search, YouTube, Display and the rest in real time is simply not a human job.
• Round-the-clock service in more languages. Klarna’s assistant ran in 23 markets and more than 35 languages, according to its own release.
• Revenue, where it is measured. McKinsey’s 2026 survey found revenue gains are most often attributed to AI use in marketing and sales, ahead of other functions.

The less flattering part of the same McKinsey survey is that only 37% of respondents attribute any EBIT impact to AI at all. Gains in a function are easier to claim than gains on the profit and loss account, and if you need to prove the difference to a finance director, measuring the ROI of AI in marketing sets out how I’d do it.

What are the risks and limits of AI in marketing?

AI creative can look fine and still sell less

This is the finding I’d put in front of any leadership team before they sign off an AI creative programme. Adam Peruta, associate professor at Syracuse University’s Newhouse School, co-led a study with Ipsos in which 3,000 US consumers watched 20 video ads for major brands: half the original human-made spots, half rebuilt entirely with AI from briefs based on the originals. People struggled to tell them apart. Yet on average the AI versions were 14% weaker on short-term sales potential and 17% weaker on long-term brand equity (HBR, September 2026).

AI did well on functional, product-led work: its Febreze and Herbal Essences ads matched or beat the human versions on short-term sales potential, and a Cheerios ad built on a strong brief came close to the original. It did badly where the idea needed a joke or real emotional precision, such as a Fiat spot that took “rebellious” at face value and produced dark neon shots of a car, while the human version turned the same “testing” idea into a gag. Peruta’s warning is that the damage doesn’t show at the point of approval, because the CFO sees lower production costs and the CMO sees faster campaign cycles. My worry is what happens next: a year later the tracker dips and we blame the category, the media mix, the new logo and probably the weather.

Consumers have views on this too. Gartner found that 49% of US consumers (307 surveyed in March 2026) think generative AI has made content quality worse, rising to 57% among Gen Z and millennials.

Automating a task without redesigning the process

Shunko and Netessine’s HBR piece, “Stop Automating Old Processes. Design New Ones Instead.”, uses Deloitte Australia as the cautionary tale: in 2025 it agreed to partially refund the Australian government after a report was found to contain fabricated references and a made-up legal quote, and a revised version disclosed that Azure OpenAI had been used. Their point is that the firm sped up the drafting without a strong enough verification gate around it, when verified findings were what the client was actually paying for. Marketing has its own versions: claims copy nobody checked, a chatbot that closes tickets quickly by not solving anything, a dashboard that improves while sales don’t.

Trust, disclosure and regulation

In the UK there is no blanket legal requirement to disclose AI use in ads. The Committee of Advertising Practice said so in May 2025, adding that its existing rules “apply regardless of how content is generated”, so the test is whether the audience would be misled without a disclosure (think deepfakes or AI influencers). Peruta’s HBR article notes that the EU AI Act brings transparency obligations for certain AI-generated or manipulated content from August 2026, and that a large majority of his participants wanted AI use disclosed. In regulated categories the bar is higher again; if you work in healthcare, the UK pharma advertising rules apply to AI output exactly as they do to anything else.

Inside the organisation, the problem is trust in agents. Thomas McKinlay (Science Says), Stefano Puntoni (Wharton) and Serkan Saka (San José State University) write in HBR that 59% of enterprise organisations say they use agentic AI but only 9% have turned that into autonomous workflows. Gartner’s separate survey of 402 CMOs points at the human side of the same gap: 80% say staff fear and anxiety is a barrier to AI experimentation, even as they expect the share of marketing work automated by AI to rise from 16% in 2026 to 36% by 2028 (Gartner, May 2026).

How to use AI in marketing strategy: where to start

This is the order I’d follow. I lead AI adoption in marketing at the charity ACTAsia, so it’s the sequence I think about most.

1. Start with one workflow. Map one process end to end (say, monthly email production or paid social creative) and mark where the value is created and where the checks are. Shunko and Netessine’s diagnostic question is a good one to pin up: where will the bottleneck move once this step gets faster?
2. Pick conventional, high-volume work first. Peruta lists resizing, versioning, localisation, first-draft copy options and straightforward product demonstrations as the execution work generative AI speeds up, and his best-performing AI ads were exactly that kind of functional work.
3. Write down the limits for your team. McKinlay, Puntoni and Saka found that telling people where an AI is likely to fail made it seem up to 14.6% more transparent and helped them work with it up to 7.2% more effectively, even though the AI itself was identical.
4. Keep a human on the send button. Their quote from Workato’s CTO, Adam Seligman, is the right rule for most marketing agents: “You can draft this email, but you can’t send it.”
5. Test AI work against human work on the same brief. Measure attention, persuasion, conversion and brand metrics rather than cost per asset, then scale what wins and shelve what doesn’t.
6. Decide on disclosure before you need it. Keep a record of where AI was used, and set a house rule on synthetic people and voices that agencies have to follow too.

For the build itself (triggers, prompts, review gates, handoffs), see how to build an AI marketing workflow. And a reasonable first prompt for step 1, if you want to try it this afternoon, looks like this:

I run [process, e.g. our monthly donor email] for a [type of organisation].
Here are the current steps, who does each one and how long it takes: [paste].
1. Identify which steps are repetitive and pattern-based, and which depend on judgement, brand knowledge or compliance sign-off.
2. For each repetitive step, suggest how AI could do a first draft and what a human should check before it moves on.
3. Tell me where the new bottleneck is likely to be, and what could go wrong if we skip the checks.
Do not suggest removing any approval step. Flag anything you are unsure about.

Which AI in marketing tools are actually worth using?

You need fewer tools than the conference expo hall suggests, and many of the useful ones are already in your stack.

• General assistants (ChatGPT, Claude, Gemini). Good for drafting, summarising research, restructuring messy data and pressure-testing a brief. Bad at facts you haven’t given them, and they will happily invent a statistic if you ask for one, so treat every number as unverified until you’ve found the source.
• Ad platform automation (Google Performance Max and its equivalents on other platforms). It handles bidding and placement at a scale humans can’t match, though it won’t tell you much about why something worked, and it is only as smart as the conversion signal you feed it.
• CRM and marketing cloud AI (for example Salesforce Marketing Cloud Next). Useful for segmentation, journey building and, increasingly, agent-led campaign set-up, with the obvious condition that it is only as good as your customer data, which for most of us is a sobering thought.
• Creative production platforms (such as the Nvidia Omniverse set-up Unilever uses). These earn their keep on consistent product imagery and localisation at volume. They are expensive to set up and suit brands with large product ranges far more than a small team with ten SKUs.

If I had to pick one thing to do this month, it would be to take the last few AI-made assets your team shipped and ask whether anyone tested them against a human-made version on the same brief. If nobody did, that test is where I’d start, before buying anything new.

FAQs

What companies use AI in marketing?

Plenty of well-known brands have published what they do. Documented examples include Unilever, which uses AI digital twins of products such as Dove and TRESemmé to create imagery; Netflix, which personalises artwork for each member; Klarna, whose AI assistant handled two-thirds of customer service chats in its first month; Coca-Cola, which rebuilt its “Holidays Are Coming” ad with generative AI; and Wayfair, which used AI to correct 2.5 million product tags.

What is an example of AI in online marketing?

Google’s Performance Max is a good one. An advertiser sets a budget, a conversion goal and a set of text, image and video assets, and Google’s AI decides which combinations to show, to whom and at what bid across YouTube, Display, Search, Discover, Gmail and Maps. It is entirely possible to use AI this way for years without ever describing it as an AI project.

Does AI-generated advertising work as well as human-made ads?

Not on average, according to a study by Adam Peruta of Syracuse University with Ipsos, published in HBR in September 2026. Across 20 ads shown to 3,000 US consumers, fully AI-made versions were 14% weaker on short-term sales potential and 17% weaker on long-term brand equity, even though people struggled to tell them apart. AI did best on simple, product-led ads.

Do you have to disclose AI use in UK advertising?

There is no blanket legal requirement in the UK, according to CAP guidance from May 2025. Existing advertising rules apply however content is made, so you should disclose when leaving AI use unmentioned would mislead people, for example with deepfakes or AI-generated influencers. A label won’t rescue an AI image that exaggerates what a cosmetic product does, which is likely to mislead anyway. EU rules on certain AI-generated content apply from August 2026.

How much of the marketing budget goes on AI?

Gartner’s 2026 CMO Spend Survey of 401 marketing leaders in North America, the UK and Europe found CMOs allocate 15.3% of their marketing budgets to AI. Only 30% said their AI capabilities were mature enough to scale, and organisations Gartner classed as AI-ready put 21.3% of their budgets into AI.

Sources

1. Gartner, “Gartner 2026 CMO Spend Survey Finds CMOs Allocate 15.3% of Marketing Budgets to AI, But Only 30% Are Ready to Scale AI Capabilities”, 11 May 2026. gartner.com
2. Gartner, “Gartner Survey Finds 49% of U.S. Consumers Say GenAI Has Made Content Quality Worse”, 9 June 2026. gartner.com
3. Gartner, “Gartner Survey Reveals Marketing Leaders Expect AI Automation of Marketing Work to Double to 36 Percent by 2028”, 11 May 2026. gartner.com
4. McKinsey & Company (Tinkoff, Van der Veken, Chui, Balakrishnan), “The state of AI in 2026: On the road to ROI”, 25 August 2026. mckinsey.com
5. Harvard Business Review (Adam Peruta), “Research: AI-Generated Ads Perform Worse Than Human-Made Ones, Even When Customers Can’t Tell Them Apart”, 3 September 2026. hbr.org
6. Harvard Business Review (Masha Shunko and Serguei Netessine), “Stop Automating Old Processes. Design New Ones Instead.”, 14 September 2026. hbr.org
7. Harvard Business Review (Thomas McKinlay, Stefano Puntoni and Serkan Saka), “To Adopt AI at Scale, Employees Need to Trust Agents”, 9 September 2026. hbr.org
8. Klarna via PR Newswire, “Klarna AI assistant handles two-thirds of customer service chats in its first month”, 27 February 2024. prnewswire.com
9. CX Dive, “Klarna changes its AI tune and again recruits humans for customer service”, 9 May 2025. customerexperiencedive.com
10. Unilever, “Unilever reinvents product shoots with digital twins and AI”, 18 March 2025. unilever.com
11. The Coca-Cola Company, “Coca-Cola Refreshes Givers of the Season, Embraces AI-Powered Storytelling in Global Holiday Campaign”, 3 November 2025. coca-colacompany.com
12. LBB Online, “Cadbury and Shah Rukh Khan Enlist the Help of AI to Promote Hundreds of Small Businesses This Diwali”, 27 October 2021. lbbonline.com
13. Netflix Technology Blog (Chandrashekar, Amat, Basilico, Jebara), “Artwork Personalization at Netflix”, 7 December 2017. netflixtechblog.com
14. Google Ads Help, “About Performance Max campaigns”, accessed 23 September 2026. support.google.com
15. Salesforce, “Salesforce Ends the Era of ‘Do-Not-Reply’ Marketing with Marketing Cloud Next”, 11 June 2025. salesforce.com
16. ASA and CAP, “Disclosure of AI in Advertising: Striking the Balance Between Creativity and Responsibility”, 29 May 2025. asa.org.uk