Marketing
Shorten the time from creative brief to launch, scale asset production, and keep brand guardrails tight.
Updated 4 days ago · 10 sources
Proof over hype
Rate any example below — it feeds which ones stay on this page.
- Enterprise
L'Oréal
L'Oréal's 10,000 marketing staff worldwide needed far more assets per launch, in more sizes and languages, without paying agency rates for every version.
L'Oréal built its own AI system, CreAItech, that generates the photos and video for each version instead of a studio reshooting it — production costs down 40%, and 50,000 marketing assets produced in roughly a year.
The takeaway The 40% saving is real, but notice where it's concentrated: repeatable, high-volume production like resizing and localization, not the one flagship idea a brand only makes once a year.
How they did it
L'Oréal built CreAItech, an in-house system that pipes together AI models from multiple providers — Seedance, Google's Gemini, Veo3 and Imagen, Adobe, Stable Diffusion and, as of this year, OpenAI — into one production workflow, so a marketer can switch models depending on what a given asset needs rather than being locked into one vendor's output. It runs on Nvidia infrastructure L'Oréal built out specifically to handle the volume once usage outgrew a single system. On the company's full-year earnings call, CEO Nicolas Hieronimus told analysts staff had used the CreAItech suite to produce 50,000 marketing assets in the roughly year since it rolled out, at 40% lower production cost than before — framing it modestly for a headline number: 'We have to learn to master the tools, but it's really a new era.' The new OpenAI deal folded into the system also gives L'Oréal access to a life-sciences reasoning model for its research side, and lets the company feed its own current product data into ChatGPT's underlying models, so AI assistants answering questions about L'Oréal products draw on the company's own information rather than whatever is floating around the open web.
- EnterpriseSilverside
Svedka
Svedka wanted a Super Bowl ad people would talk about for how it was made: rebuilt with AI instead of a traditional animation studio, using its long-running robot mascot Fembot.
The AI learned to reproduce Fembot's exact expressions and movements well enough for her to share the screen with a new co-star, Brobot, in a national Super Bowl spot.
The takeaway The 'AI-generated' headline undersells the labor: four months of iteration, and humans still wrote the storyline. AI replaced the animation studio's work, not the creative idea.
How they did it
Working with the studio Silverside, Svedka rebuilt its long-running robot mascot, Fembot, and introduced a new companion character, Brobot, dancing together at a party for what the brand calls the first primarily AI-generated national Super Bowl commercial. 'Primarily' is doing real work in that description: humans still wrote the storyline and creative direction, and reconstructing Fembot meant training the underlying AI systems to replicate her specific facial expressions and body movements convincingly enough to hold up on the biggest screen of the year, not generating a generic robot from scratch. Getting a consistent, ownable character to move and emote correctly took about four months of iteration, not a single prompt.
- Enterprise
Booking Holdings
Booking Holdings wanted its brand to actually appear in the answers AI assistants give when someone asks about travel, rather than getting summarized out of existence by a chatbot that never sends a click back to its site.
The takeaway The skill AI search rewards is the opposite of what social media trained marketers to do for a decade: depth beats punchy, because a short one-liner gives an AI assistant nothing to quote. Booking hasn't published traffic or citation numbers from the shift, but the fix itself is concrete — write longer, more detailed pages, not shorter ones.
How they did it
Chief communications officer Leslie Cafferty said the fix cuts against a decade of social-media-trained instinct: AI assistants need real substance to quote, not a punchy one-liner. Booking is adding meaningfully more detail to its own web pages so there's something worth citing when an AI assistant answers a travel question, moving past the 'short and pithy' copy that wins on Instagram but gives a language model nothing to work with. Other marketers making the same shift described it in similar terms: Getty Images' Susan Nomecos summed up the underlying risk as 'garbage in, garbage out' for any brand feeding thin content into these systems, and JPMorgan Chase took a parallel approach on the listening side, training AI to read 'nuanced emoji, sarcasm, and slang' in social conversation well enough to catch sentiment that traditional keyword monitoring misses entirely.
- Learn from thisEnterprise
McDonald's Netherlands
McDonald's Netherlands wanted a Christmas ad that would stand out on social media.
Pulled from air after just three days online, following viral backlash.
So: Show any AI-made ad to a real sample of your own customers before it goes live, and agree in advance, in writing, who has the authority to pull it fast if it lands wrong.
The takeaway AI is strong behind the scenes on production. It's much weaker at the emotional judgment a holiday ad actually depends on, and nobody at McDonald's caught the tone problem before the internet did — a national survey backs that up: 65% of US consumers say AI-generated ads make them uncomfortable, and over 30% say knowing an ad is AI-made makes them less likely to buy.
How they did it
The company posted a 45-second AI-generated ad to its Dutch YouTube channel: a string of AI-generated characters stumbling through Christmas mishaps before finding refuge from 'the madness, the lights, and the cheer' at McDonald's, set to a parody of "It's the Most Wonderful Time of the Year" that swapped 'wonderful' for 'terrible.' Viewers found the stitched-together AI clips uncanny and emotionally flat rather than festive, and the message itself read as cynical against a season built around warmth and togetherness at home, pitching McDonald's as an escape from Christmas rather than part of it. Commenters had a one-word verdict for the whole thing: 'AI slop.' McDonald's pulled the ad just three days after posting it and told the BBC the episode was 'an important learning' about the effective use of AI.
- Learn from thisEnterpriseSilverside AI
Coca-Cola
Coca-Cola made another AI-generated version of its classic 'Holidays Are Coming' Christmas ad for 2025 — a year after its first AI Christmas ad was met with public outrage and mockery — and used AI again anyway.
So: Budget human review capacity that scales with how much content the AI generates, not with how much time you saved on the shoot — someone still has to check all of it. And test public reaction directly before you repeat a creative approach that drew backlash once already, rather than assuming a repeat performance is safer.
The takeaway AI cut the production timeline dramatically — five people made this in about a month, versus a full production crew and schedule for a traditional shoot — but it didn't cut the review workload, it moved it: roughly 100 people still had to check 70,000-plus generated clips before anything could air. If you're piloting AI-generated creative, budget human review capacity that scales with how much the AI generates, not less of it. And don't assume using the same approach again will avoid the same public reaction as last time — measure that directly instead of assuming it.
How they did it
Coca-Cola worked with AI production studio Silverside AI to remake the ad. A team of just five 'AI specialists' generated and refined more than 70,000 video clips over about a month to get the footage, with roughly 100 people across the production reviewing that output before anything aired — AI didn't remove the human review step, it multiplied how much there was to review. The finished ad drew the same kind of public criticism and mockery as the original 2024 version.
Try this today
Turn one brief into six drafts
20 minSix subject lines and three ad drafts that follow your own voice rules, not generic AI copy.
Copy the prompt
You are a senior copywriter. I will give you our campaign brief and three pieces of copy we have already published. Step one: read the three examples and write five specific rules about how we write — sentence length, whether we ask questions, words we never use. Step two: using those rules, write six email subject lines and three 40-word ad versions. Use only claims that appear in the brief. If you wanted to add a claim that is not in the brief, list it separately under "claims I did not use". Brief: [PASTE YOUR BRIEF] Examples: [PASTE THREE PIECES OF PUBLISHED COPY]
One check first. Check every claim against the brief and your approved wording. The five rules are the model's guess, so a senior writer signs those off.
Find the words your customers use
15 minThe main themes in your own review data, plus the exact phrases customers repeat back.
Copy the prompt
You are a customer researcher. Below are customer reviews, survey answers, or support messages. Do four things. One: group them into no more than eight themes and say how many comments sit in each. Two: for each theme, quote three real lines word for word. Three: list the phrases customers use for our product, and separately the phrases they use for the problem it solves. Four: name anything they complain about that our current marketing does not answer. Do not rewrite them in marketing language. Use their words. Comments: [PASTE THE COMMENTS]
One check first. Strip names and emails before you paste. The counts only cover what you pasted, so check them against a full export.
Prep the questions your CFO will ask
25 minYou walk into the budget meeting already holding answers to the ten hardest questions.
Copy the prompt
You are a sceptical chief financial officer reviewing a marketing budget request. Below is my campaign plan. Ask me the ten hardest questions you would ask in the meeting, hardest first. For each one, say what evidence would answer it and where a marketing team usually finds that evidence. Then mark every number in my plan you would not accept without a source. Do not invent benchmarks or industry averages. If you do not know a figure, say that instead. Plan: [PASTE YOUR CAMPAIGN PLAN OR BUDGET REQUEST]
One check first. The model does not know your finance team or last year's results. Treat the list as a rehearsal and bring your own numbers.
Specialized tools, and what to ask vendors
| Tool | What it does | Setup | Best fit |
|---|---|---|---|
| Jasperfrom $59 per seat, billed annually | Writes campaign copy from a brief and holds your brand voice (how your writing should sound) in every draft. | Under a weekIT sign-off | You ship a lot of channel copy and brand consistency keeps failing at the edges.Skip it ifYou need one big idea, not volume. Jasper is a production tool, and the blank page is still yours. |
| Adobebundled into Creative Cloud plans | Makes images and video from a text prompt, with more than thirty models inside one creative studio. | Under a week | Your team already lives in Photoshop and you need many sizes of the same asset.Skip it ifHero work with real people. Faces and hands are where AI imagery still gets caught in public. |
| HubSpotbilled per credit, not per seat | An assistant and set of agents inside HubSpot that draft content, research prospects, and answer questions from your data. | Under a weekIT sign-off | HubSpot is already your CRM (the system that holds your customer records) and you want the short path.Skip it ifYou want the strongest writing tool on the market. This one wins on convenience, not on output quality. |
| Helenafrom $39 per month, per Enrich Labs | A marketing agent that can turn your site, analytics and brand information into SEO drafts, social posts, email campaigns and a daily performance briefing.Try Helena for marketingReferral link | Under a weekIT sign-off | You need consistent marketing execution and can name one person who will approve public posts, email sends and ad spend.Skip it ifYou want an agent to set strategy or publish and spend from your accounts with no human review. Start with drafts and reporting first. |
| Persadoenterprise, quoted | Generates and scores marketing language, then checks it against the rules that govern regulated ads before you send. | MonthsIT sign-off | You market a bank, insurer, or health product, and legal review is your real bottleneck.Skip it ifSmall budgets and low send volume. The scoring needs a lot of traffic before it earns its price. |
Questions to ask before you buy
Jasper — 7 questions to ask them
Also used by. Ulta Beauty, Prudential, Hilton and Sanofi, per the logos on its own site
- Who owns the copy Jasper produces, and does that change between the Pro plan and Business?
- How many approved samples do you need before the brand voice stops sounding generic?
- Are my brand guidelines, product data and customer copy used to train anything you sell to others?
- Show me a customer in a regulated category and how their legal team signs off on Jasper output.
- At $59 a seat, what does full rollout cost once every marketer needs a login?
- Which features are Business-only, and which of the ones you just demoed are on that list?
- Which customers left last year, and was it price, output quality, or the voice going off-brand?
Adobe — 7 questions to ask them
- Which plan includes IP indemnification (they pay the lawyers if we get sued), and what voids it?
- What were these models trained on, and will you put that answer in the contract?
- Can I generate a recognisable human face for a paid ad, and would you defend that?
- How do I stop a junior designer publishing an AI image that nobody reviewed?
- What does this cost on top of the Creative Cloud seats I already pay for?
- How do generation limits work, and what happens in a month when we run a big campaign?
- Which brand my size ran a paid campaign on Firefly assets, and what did their legal team reject?
HubSpot — 7 questions to ask them
- How do credits (prepaid units an agent spends each time it works) convert into real spend for me?
- What does an agent do with a contact record that is wrong or years out of date?
- Which agents are marketing agents, and which ones are really sales or support tools?
- Is anything I generate here used to improve models that other customers benefit from?
- If I stop paying, do I keep the content and the data the agents produced?
- What is the cheapest tier where the marketing AI features are actually included?
- Which customer my size measured a real lift from the agents, and can I talk to them?
Helena — 6 questions to ask them
- Which actions are drafts, and which ones can publish or spend from our accounts without another approval?
- Can we begin with read-only access to analytics and a draft-only publishing workflow?
- How does the agent separate a recommendation from a claim pulled from our own campaign data?
- What happens when the agent reaches a login challenge, a platform policy change or an unfamiliar page?
- Show me the complete log for one customer task, including the work it did not complete.
- What data from our analytics, ads and customer lists is retained, and can we delete it on exit?
Persado — 7 questions to ask them
Also used by. Large US banks and insurers, per the customer logos on its own site
- Which regulatory frameworks do you check against, and who keeps that list current?
- If your copy passes your compliance check and my regulator disagrees, who carries that?
- How much send volume do I need before your scoring beats a plain A/B test?
- Does the language you generate sound like my brand, or like every bank you serve?
- Show me the control group behind one lift number in your published case studies.
- What is the first-year cost including setup, and what does my team have to staff?
- Which customers did not renew, and what did they say was missing?
What everyone is asking
- NewJul 2026
Will AI actually cut my content production costs?
L'Oréal told analysts its in-house AI system cut production costs 40% across 50,000 assets. Svedka's Super Bowl ad still took about four months to make.
What to watch for
The savings show up in repeat work, like resizing and localising. One-off flagship work still takes months.
Also worth a look. Volume content production, One-off hero campaigns
- NewJul 2026
Does AI content hurt your search rankings?
Google says how content was made is not the issue. Its scaled content abuse rule — aimed at mass pages built to game search — is what thin AI pages published in bulk break.
What to watch for
Google updated this guidance in December 2025. The penalty targets thin pages made in bulk, not the tool.
Also worth a look. Publishing AI drafts with real editing, Publishing AI pages at volume
- NewJul 2026
Do customers mind if an ad was made with AI?
Enough of them do. eMarketer reported 65% of US consumers are uncomfortable with AI-generated ads. McDonald's pulled a Dutch Christmas ad after backlash.
What to watch for
Coca-Cola took the same criticism and ran the ad anyway, saying it tested better than almost anything before it.
Also worth a look. McDonald's Netherlands, Coca-Cola
- NewJul 2026
How do I get my brand into ChatGPT answers?
Write pages with enough detail to be quoted. Booking Holdings is adding depth to its sites. Vendors call this generative engine optimisation — getting cited by AI assistants.
What to watch for
Track mentions and citations, not clicks. Digiday reported the median company appears in 16% of relevant AI answers.
Also worth a look. Rewriting site content for depth, Tracking mentions and citations
- NewJul 2026
Where does AI pay off if I am not making ads?
Quieter jobs pay first. JPMorgan Chase uses AI to read tone and slang in online conversation. Nestlé builds digital copies of products to make e-commerce images.
What to watch for
Neither company published a number. Both built the capability in-house rather than buying a finished product.
Also worth a look. Consumer insight and listening, Product imagery for e-commerce
Worth following
Paul Roetzer — Marketing AI Institute
Founder of Marketing AI Institute and co-host of a weekly AI news podcast for marketers.
weekly · podcast and newsletter
Why them
Weekly news plus courses, written for marketers rather than engineers.
Marketing Brew
Weekday marketing news briefing from Morning Brew, written for people who run marketing.
weekdays · newsletter
Why them
Short, readable, and it quotes named marketers at named companies about what they built.
Rand Fishkin — SparkToro
Founder of SparkToro and Moz, writing about audience research and how search behaviour shifts.
monthly · blog
Why them
He publishes the underlying click and search data rather than summarising someone else's.
Marketing Dive
Daily marketing trade news from Informa TechTarget, covering campaigns, agencies and ad technology.
daily · publication
Why them
Original reporting on named brands, with the numbers companies actually disclosed.
Digiday
Trade publication covering media, agencies, and where advertising money is moving.
daily · publication
Why them
It gets marketers on the record about AI spending and results before the case studies exist.
Ann Handley
Marketing writer, co-founder of MarketingProfs, and author of books on business writing.
fortnightly · newsletter
Why them
Her fortnightly letter is about writing well, which is what AI drafts still need.