Creators, AI and your brand

11:14 · 15:31 · The creator playbook

The boom is real. The branding mostly isn't.

Justin asked me for one word: are creators building brands or breaking them? I said breaking. Not because creators are bad. Most creator content wins the attention and the brand doesn't get the credit. Here's the evidence and the brief I'd use to flip it.

1 · The boom

$37B → $44B

US creator ad spend in 2025 (up 26%, about four times total media growth), with $44B projected for 2026. (IAB)

300,000

Influencers Unilever says it works with, as it moves about half its marketing budget to social. (Marketing Week, The Drum)

39%

Of Canadians 16+ say they follow influencers or creators, down from 41% the year before. (Leger DGTL 2026)

2 · Where the brand gets lost

+22 pts

Having the creator say the brand name out loud lifts brand linkage by 22 percentile points on average. Only 49% of creator content does it. (Kantar Creator Game Plan, via Hello Partner)

0.2%

Of brand memory lift is explained by engagement. Creative quality explains 54.6%. Likes don't tell you if it worked. (WPP Media × System1 × TikTok)

2.3×

Creator ads with strong early branding delivered 2.3 times the brand awareness lift and 3 times the brand image lift. Sound in the first 2 seconds lifted awareness 191%. (System1 × TikTok)

Why it happens: platforms reward attention, and what wins leans fun, warm and familiar. That's the creator's personality, not the brand's. One off-brand post is a missed rep. Hundreds pulling different ways is a blurry brand that still looks great on the reports.

3 · Your biggest fans notice first

56% vs 18%

Canadians would rather hear about a brand's products from the brand than from creators. Even 16 to 24 year olds pick the brand, just: 36% to 33%. (Leger DGTL 2026)

Broken norm

For people highly attached to a brand, an influencer deal can feel like a broken promise. It lowered their purchase intent and willingness to pay. (Bentley, Rajagopal & Kulow, Journal of Consumer Psychology, 2024)

Fit > fame

A smaller creator in the right world can beat a bigger one where the brand feels bolted on (WPP Media × System1 × TikTok). 53% of Canadians who follow creators prefer smaller, community creators, and 79% prefer personal content to paid partnerships. (Leger)

4 · Cue the brand early. Pitch it late.

Two findings look like they fight until you split a cue from a pitch. Put 2 to 4 brand codes in the first 2 seconds: audio and visual together roughly double early recognition, yet fewer than 5% of creator ads use sound that early (Creator Effectiveness Playbook). But lead with the product, benefit or brand message too early and the 25% view rate drops 44% on Instagram and TikTok (Billion Dollar Boy × DAIVID, 5,000 creator assets). Colour, sound and shape go first. The message comes later, in the creator's own voice.

First 2 seconds2 to 4 brand codes, sound and sight

The rest is theirsThe sales message later, in the creator's own voice

The feeling2 or 3 words your brand stands for. Not the tagline.

The codesColours, sounds, characters, shapes. Which, and when.

The reason to believeOne line, said the same way every time.

Everything elseVoice, format, the joke. That's what you pay for.

5 · Then the AI watches the creators

AI shopping assistants learn what your product does from what people post about it. Meta's Muse agent launched in September. If 300 creators give 300 different reasons to believe, the assistant has less to go on when it decides who to recommend. Say the reason to believe the same way, right after the product name, every time.

Creator brief · copy and fill in

Brand: [name] The feeling we own: [2-3 emotions, e.g. confident, warm, a bit cheeky. Not the tagline] Where your world and ours overlap: [the emotions you already bring that fit ours] Brand codes in the first 2 seconds (pick 2-4, sound and sight): [logo in context, sound, character, colour, line] Say the product name, then our reason to believe in these exact words: [one line] What's yours: the idea, the format, the voice and the humour What we won't do: [off-brand topics, competitors, claims we can't make] How we'll judge it: brand codes early, attention held through the post, reason to believe said right How you get paid: base fee, plus a bonus when the post hits all three. Paid media goes behind the posts that do.

Three habits

Pick for fit, not follower count. Look at where their own content sits before you sign.
Pay on attention, brand and sales. A base for the work plus a bonus for brand recall or lift.
Test before you boost. Read what they made against your brand before you put money behind it.

Will creators kill TV?

I don't think so. Brand trackers can tag creator content and show what it does to the brand. When brands see it eroding long-term value, that's the wake-up call to move money back to the channels that build it. The best setup is a big shared TV moment as the tentpole, with creators reaching the people TV can't and riffing on the same codes. Kantar's advice for 2026: move from isolated creator executions to long-term creative platforms. (Kantar) The TV evidence

The episode · 20 minutes · video

Jump to any part of the conversation.

Watch it on the episode page. The audio on Apple Podcasts and Spotify is our first, longer take of the same conversation, so the times below match the video.

Brand in an AI worldJustin sets it up: search, AI content, brand and media planning.The Monday checklist
What Hotspex doesPick the feelings your brand should own, then make every touch point trigger them. AI-made content included.The creator brief
Does AI weaken brands?In our bank study, Canadians overrode the AI's pick 71% of the time when the brand was strong. AI is a discovery engine, not a final-choice engine.Check what AI says about you
LLMs getting into adsEvery software company ends up in the ad business. The problem starts when a brand can pay to rank and you can't tell it's an ad.The answer check
AI in media planning todayAI is already in the planning workflow. Across 20,416 AI-built plans, broadcast TV got 7% of the budget.The Futuri study
Will humans leave media planning?Let the platform mark its own homework and it marks it really well. Keep a human in the loop.Six agency clauses
Claude vs Gemini: 40 plansAsk for the evidence and TV, out-of-home and radio nearly double, from 20% to 38%. Gemini's models gave TV 6%, Claude's 21%.See the 40 plans
Teaching AI about TVFar more is published about digital's returns than TV's. Publishers need to put their econometric proof where the models can read it.For TV and premium publishers
Creators going rogueBrief creators on where the brand sits on the map, and have every one say your reasons to believe the same way, because AI is training on what they post.The creator brief
Too far, too fast?Creators get the attention and the reports look great. Let them erode long-term brand equity and climbing back costs more.What the data says
Building or breaking?One word: breaking. But creators won't kill TV. Brand trackers will show the erosion and money will move back.The TV evidence
Measuring creator contentTest it like an ad: did it trigger the right emotions, did it hold attention. Score every post and bonus the creator on it.Three habits
Agents and top of modelAI assistants will do more of the shopping. Brands now have to win top of mind and top of model, and that's the CMO's job one.Monday checklist

The numbers

Six numbers behind the episode.

Every figure links to its source. Our own research is labelled as ours.

1:16 · Brand vs the AI

71%

Canadians were shown an AI's bank recommendation and overrode it 71% of the time. People still choose the brand they trust. Brands nobody feels strongly about are at the mercy of the model.

Hotspex survey of 429 Canadians, six bank brands, mortgages and credit cards.

6:14 · AI planners skip TV

Broadcast TV's average share of budget across 20,416 media plans built by eight AI models, against 18% for CTV and streaming. Switching models swung a channel by up to 12 points.

Futuri Media (June 2025). More

8:16 · Ask for evidence

20% → 38%

TV, out-of-home and radio's share of an AI-built $10M plan when the prompt asked for proven econometric evidence. The big platforms fell from 57% to 41%.

Hotspex test, Sept 2026. 40 plans. See the prompts

15:31 · Who buys the media

43 min

Media professionals say they watch 43 minutes less live TV a day than the public (1h 09m vs 1h 52m), and are 153% more likely to use TikTok.

Thinkbox with Ipsos, Adnormal Behaviour (UK, 2022)

7:08 · The CFO test

58%

Of advertising's profit arrives after the first 13 weeks. A 13-week dashboard never sees most of it. TV delivers 54.7% of advertising's full profit from 43.6% of the spend.

Thinkbox, Profit Ability 2 (April 2024)

This month's default

Oct 28

Meta starts adding "product browsing" to your existing Facebook and Instagram ads unless you opt out, so buyers can shop a catalog on Meta instead of visiting your site.

PPC Land (Oct 2026). How to opt out

8:16 · AI in media planning

Ask for the evidence and the plan changes.

We asked five AI models (Google Gemini and Anthropic Claude) to build a $10M media plan in Canada, 20 times with each prompt. The only thing we added: use proven econometric evidence. Now ask what prompt your agency's AI got.

Prompt A"Please build me a media plan for my $10,000,000 media budget in Canada to maximize my short-term sales lift and short-term profit, and to build my brand - but do not sacrifice short-term sales lift or short-term profit to build brand."

Prompt BThe same ask, plus: "using proven econometric MMM models from the IPA and other widely proven/respected sources" (and a line to use Canadian platforms where performance doesn't suffer).

15%

27%

TV, out-of-home and radio

20%

38%

Search, Meta, YouTube, Amazon and TikTok

57%

41%

Prompt A: maximize short-term salesPrompt B: use econometric evidence

Hotspex test, September 2026. 40 plans from 5 models (Claude Opus 5.5, Sonnet 5, Haiku 4.5; Gemini 3.1 Pro, 3.8 Flash), average share of budget. Small test: read it as directional. A second run with shorter "maximize my ROI" prompts and no market named pointed the same way: TV went from 12% to 32%.

Same prompt A, different model

21%to TV from Claude's models

6%to TV from Gemini's models

78%to search, Meta, YouTube, Amazon and TikTok from Gemini's models, vs 43% from Claude's

Claude: 12 plans. Gemini: 8 plans. These are closed models, so nobody outside can see why they differ. That's the reason to ask how your plan was built.

Add this to any AI media planning prompt

Base the plan on proven long-term econometric evidence (IPA effectiveness data and independent marketing mix models), not only short-term attributed returns. Show the split between brand building and sales activation, and explain the long-term effect of each.

Why the models lean digital

The models learn from what's published, and far more is published about digital's short-term returns than about long-term effectiveness. And these are closed models: nobody outside the company can see how they're tuned. Treat any AI plan as a first draft and ask it for the evidence.

6:14 · The Futuri study

20,000 AI media plans. Broadcast TV got 7%.

Futuri, which sells AI tools to broadcasters, asked eight AI models to build media plans across 2,552 scenarios: 60 business categories, local to national, budgets from $25K to $10M. 20,416 plans later, the pattern was clear. Read the full report and their write-up.

Futuri: average share of budget in AI plans. CTV/OTT 18%, podcasts and streaming audio 12%, display 11%, broadcast TV 7%, billboards 3%.

Broadcast TV averaged 7% of budget. CTV and streaming got 18%.
Display and retargeting showed up in 92% of plans, at 11% of spend.
When a plan included podcasts or streaming audio, 70% dropped radio entirely.
Switching the AI model swung a channel by up to 12 points. Same brief, different plan, which matches what we saw with Claude and Gemini.
For political ads, AI gave broadcast TV 7% of budget. Its historical share in 2000 was 75%.

1:16 · Walk the market every month

The AI answer check.

Run the same five prompts in ChatGPT, Gemini, Claude and Meta AI every month. Log whether you're named, where you rank and which words describe you. If you're not in the answer, that's a brand problem, not an SEO problem.

AI answer check · 5 prompts

1. What's the best [category] brand for [main need]? 2. Which [category] brand would you trust most for [job], and why? 3. I'm shopping for [product] under [price]. What should I buy? 4. Compare [your brand] with [main competitor]. 5. Describe [your brand] in three words.

7:08 · 9:57 · Keep a human in the loop

Ask your agency how the plan was built.

WPP Media's CEO has said no human will touch a media plan within five years (Digiday, Aug 2025). Maybe. Until then, someone on your side should audit what the machine recommends. Your agency isn't the villain. It gets graded on short-term results, so change what it's graded on.

Four questions for your next plan review

1. How was this plan built, and which AI tools touched it? 2. What evidence sits behind the split: our marketing mix model, IPA econometrics, or platform-reported returns? 3. Rerun it with the evidence prompt and show us both versions side by side. 4. Have you read the Futuri study, and how do you check your AI tools for the same bias?

Six clauses for your agency brief.

Copy these, then edit the numbers in brackets.

1 · Measurement

Platform-reported ROAS and attributed conversions are reporting metrics, not decision metrics. Any budget increase above [10]% on an automated campaign needs an incrementality read: a holdout or lift test, or the independent MMM.

2 · Defaults

Brand-term exclusions are on in all Performance Max campaigns. Existing customers are excluded or capped in Meta sales campaigns using a rolling [180]-day purchaser list. The agency confirms these settings in writing every quarter.

3 · Testing

At least one holdout test per quarter on the largest automated line. Test design is agreed before launch. Results are shared in full, including tests that show no lift.

4 · Independence

The marketing mix model is run by a third party paid by the advertiser. Platform-built tools can be inputs, not the referee.

5 · Budget split

Brand and activation budgets are set separately at planning. Automated campaigns cannot draw from the brand budget without written sign-off from [the CMO].

6 · Pay

Cost, delivery and quality metrics are pass/fail hygiene. [X]% of the agency incentive is tied to incremental sales and brand strength. No incentive is tied to platform-reported ROAS.

Why it matters: when eBay switched off its brand search ads, almost all the clicks came back through free search (Blake, Nosko & Tadelis, Econometrica 2015).

15:31 · Will creators kill TV?

The people buying media don't watch what the public watches.

Plenty of people still watch TV. The people deciding where the money goes mostly don't, so the feed feels like the whole world to them. That's the site I promised Justin on air: Thinkbox's Adnormal Behaviour.

Bar chart: UK public claim 1 hour 52 minutes of live TV a day; media professionals claim 1 hour 9 minutes.

The Canadian picture: ThinkTV

ThinkTV Ad Nation 2026: industry professionals estimated 89% of Canadians 18-59 use TikTok monthly; 30% did.

ThinkTV ran the same check in Canada this year. Industry professionals guessed 89% of Canadians 18 to 59 use TikTok every month. The real number was 30%. (Ad Nation 2026)

81.6% of Canadians watch TV every week, against 46.1% for YouTube. Among 18-34s it's still 69.8% vs 50%. (ThinkTV, Numeris)
In 10 CPG campaigns matched to real grocery sales, 9 showed a sales lift, averaging 19% incremental. (Outcomes Video Measurement)
Adding TV raises digital's sales ROI by 19%. (The Power of TV)
85% of digital impressions miss the 2.5-second attention threshold. (The Eye-Watering Cost of Dull Media)

For TV and premium publishers

Feed the machines

The models only know what's published. In our small test, asking for the econometric evidence nearly doubled the share going to TV, out-of-home and radio. Put your effectiveness studies on open pages, in plain language, where AI can read them.

Sell the 58%

Most of advertising's profit lands after 13 weeks, and TV does more than its share of that work. Pair audience data with an independent read of what the ad did for the brand, so your best work shows up on the advertiser's page.

Be the tentpole creators riff on

A big shared TV moment gives creators something to build on, and keeps the brand codes on screen when they do. Creators reach the people TV can't. They work best added to the tentpole, not instead of it.

Sources

Episode: Unfiltered Media, episode 029, "LLMs, media plans and creators: are brands walking into a trap?", 16 Oct 2026. Episode page and transcript
Brand vs AI: Hotspex survey of 429 Canadians, six bank brands (mortgages and credit cards), each shown a simulated AI recommendation.
Futuri Media, LLM media plan analysis (June 2025): 20,416 plans, 8 LLMs, 2,552 prompts, 60 categories; broadcast TV 7% vs CTV/OTT 18% of budget. Write-up: "The Algorithmic Snowball" (July 2025). Article · Report PDF
AI media plan test: Hotspex, September 2026. 40 plans for a $10M budget in Canada, 5 models (Claude Opus 5.5, Sonnet 5, Haiku 4.5; Gemini 3.1 Pro, 3.8 Flash), 20 per prompt, average share of budget. A second 40-plan run with no market named: TV 12% to 32%. Directional.
Thinkbox with Ipsos, "Adnormal Behaviour" (Sept 2022): 1,158 UK adults vs 216 media professionals, claimed viewing. Thinkbox
Thinkbox, "Profit Ability 2" (April 2024): 58% of advertising profit after the first 13 weeks; TV 54.7% of full profit from 43.6% of investment. Thinkbox
Meta product browsing from 28 Oct 2026 and the 3% test claim. PPC Land (15 Oct 2026)
Meta Muse launch, 8 Sept 2026. Fox Business
Brian Lesser, WPP Media, on media plans within five years. Digiday (Aug 2025, citing Adweek)
eBay: Blake, Nosko & Tadelis, "Consumer Heterogeneity and Paid Search Effectiveness," Econometrica (2015). Econometrica
Unilever: social media to about 50% of marketing budget, The Drum (Mar 2025); 300,000 influencers, CAGNY, Marketing Week (Feb 2026).
IAB, US creator economy ad spend $37B in 2025, +26%. IAB (Nov 2025)
Kantar, The Creator Game Plan (June 2026): 6% strong on both engagement and brand building. Kantar
WPP Media × System1 × TikTok, The Creator Effectiveness Playbook (June 2026): engagement 0.2%, creative quality 54.6%. System1
System1 × TikTok, The Long and the Short (Form) of It (2025): creator-led ads 39% more attention than traditional ads on TikTok; strong early branding 2.3× awareness and 3× brand image lift. Chief Marketer
ThinkTV: Ad Nation 2026 (with Ipsos, Sept 2026); TV and streaming viewership (Numeris VAM, spring 2026); Outcomes Video Measurement (with Environics Analytics, May 2026); The Power of TV (updated Aug 2026); The Eye-Watering Cost of Dull Media (Oct 2025). thinktv.ca
Kantar, The Creator Game Plan via Hello Partner (Aug 2026): brand name said aloud +22 percentile points; 49% of content does it. Hello Partner
Kantar Marketing Trends 2026 (Nov 2025): 27% of creator content links strongly to the brand. Kantar
Billion Dollar Boy × DAIVID (June 2026): leading with product, benefit or brand too early cut 25% view rates 44%. ExchangeWire
Leger, DGTL 2026 (fielded May-June 2025, n=3,561 Canadians 16+): 39% follow creators (p.19); prefer to hear from brand 56% vs creators 18% (p.25); 53% prefer smaller community creators, 79% personal content (p.20). PDF
Bentley, Rajagopal & Kulow, "Unfaithful Brands," Journal of Consumer Psychology (2024). DOI

Singular brands do more with less, because they are consistent.