AI for Marketers in 2026: Briefs, Ad Variants, UTMs and the Legal Line
Campaign briefs, ad variants, repurposing, UTM cleanup, approvals, AI Overviews: where AI helps marketers in 2026, and the FTC rules that still apply.
AI can draft campaign briefs, cut one webinar into clips and emails, produce ad variants, tidy UTM sheets and summarize approval threads. It can't choose your positioning, settle attribution, substantiate a claim or give you a copyright in its raw output. For marketers in 2026, the real gains sit in production and admin.
Your week rarely goes wrong on the big idea. It goes wrong when legal sits on the landing page, the agency sends last round's cut, and finance asks why the ad platform and the CRM disagree. Below is that work task by task, with prices as vendors list them in November 2026, then the US Federal Trade Commission (FTC), Copyright Office and ad-platform rules a faster workflow doesn't change.
Where a marketer's hours go
A marketer's hours go less to campaign ideas than to the coordination around them. The US Bureau of Labor Statistics lists planning campaigns, negotiating contracts, reviewing how displays and websites look, and discussing budgets and creative direction among the duties of marketing managers. O*NET adds working with legal staff on copyright and royalty sharing. In practice, that looks like:
- Rewriting a stakeholder's three-line request into a campaign brief an agency can quote against.
- Moving everything on the content calendar when a launch date slips.
- Finding out why half of last month's links show up as (not set).
- Keeping track of which freelancer owes which file, and who approved which version.
- Getting a webinar's follow-up out while attendees still remember it.
Where AI fits across a marketing workload, with tools
AI does best on marketing tasks with a clear source and a human reviewer: a transcript to cut up, a sheet to clean, a brief to fill in. The tools below are examples described from each vendor's own pages, not a ranking.
Task | AI's part | Tools to look at |
|---|---|---|
Campaign briefs | Turn notes and past campaigns into a first draft in your brand voice, with gaps flagged as questions | Brand-trained writing tools. Jasper Pro lists $59 per seat on annual billing, $69 paid monthly, with 2 brand voices and 3 audiences |
Content calendars | Suggest post ideas and rewrite drafts per channel. Dates and priorities stay yours | Social schedulers. Buffer includes a visual calendar and its AI Assistant on every plan, free for up to 3 channels |
Repurposing | Find clips in a long video, write show notes and draft social posts from it | Descript lists Create Clips from its Hobbyist plan at $16 or $24 a month, by billing period. OpusClip starts at $15 a month |
Paid ad variants | Generate text, image and video assets from a prompt, ready to refine and test | Inside the ad platforms, such as Asset Studio in Google Ads |
UTM and channel reporting | Spot naming errors, pull channels into one view and draft the commentary | Data hubs. Funnel lists 117 connectors and dashboards from $300 monthly, billed annually |
Stakeholder approvals | Check a file against your rules before a person sees it | Filestage puts review agents for forbidden terms, required logos and compliance documents on its Business plan, listed at $329 a month |
AI search visibility | Track how AI answers mention your brand and which sources they cite | Ahrefs Brand Radar covers AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot |
Webinar follow-up | Summarize the session, then draft the recap email and clips | Zoom's Webinar Summary needs the Webinars add-on and has to be turned on in settings |
Creative variants for paid ads
AI makes ad variants cheap, which moves the bottleneck to review and testing. Google's Asset Studio generates text, images and video from one prompt box, but a hundred versions of a weak offer still have a weak offer. Decide what each test is meant to teach you before generating anything, and check every version for claims the original didn't make. Meta's terms say you are "solely responsible for evaluating the Output" before running it. If you sell through Google Merchant Center, AI-generated images need IPTC metadata marking them as made by a trained algorithm.
UTM and attribution reporting
AI is good at finding UTM mess and poor at telling you which channel deserves the credit. Paste a month of tagged links into an assistant and ask for every variation of source, medium and campaign. Those variations cost you data: Google Analytics treats utm_source=google and utm_source=Google as different values, and missing tags show as (not set). Ask too about links that pass through redirects or shorteners, which can lose their tags and land in (direct) / (none).
Attribution is a choice of model, and no assistant can make the choice for you. Google Analytics 4 offers data-driven, paid and organic last click, and Google paid channels last click, with a side-by-side report under Advertising. That choice is one reason an ad platform and your CRM can report different numbers, so state the model behind each before you compare them. When the CFO asks why a channel moved, have AI draft the note, then check each number against the report yourself. For the wider picture, from first click to closed deal, see sales and marketing analytics.
SEO and AI Overviews
Google says AI Overviews need no special optimization, and that mass-produced AI pages can count as spam. Its guidance states there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for a snippet. Generating many pages without adding value may violate its scaled content abuse policy, and Google asks you to fact-check AI output, including titles, meta descriptions and alt text. So use AI for outlines and refreshes, and be wary of any plan that turns one article into 400 location pages.
You can now measure part of this. Search Console's generative AI reports reached all sites in August 2026, but they show impressions only, so they can't tell you how many clicks an AI Overview sent you.
Agency and freelancer management
With agencies and freelancers, AI helps with the paper trail, and the accountability stays with you. It can compare a statement of work against what was delivered, or summarize a long thread into who owes what by when. It should not be your check on whether a freelancer used AI. In 2025 the FTC said Workado's claim of 98 percent accuracy for its AI content detector was unsupported, and that testing found 53 percent on general content. Ask for disclosure in the contract instead.
Tell us before delivery if any part of this work was generated with AI, and name which part.
Do not put our unreleased plans or customer data into AI tools we have not approved.
Do not use AI to create reviews, testimonials, people or results that aren't real.
Substantiation for every factual claim comes with the first draft.What AI doesn't change: the rules marketers answer for
US advertising law applies to AI-made marketing the same way it applies to anything else. The FTC said when it launched Operation AI Comply that there is no AI exemption from existing laws. This is general information, so take specific campaigns to your counsel.
- Fake reviews. The FTC's rule covers reviews by someone who doesn't exist, such as AI-generated fake reviews. Courts can impose civil penalties for knowing violations, and marketers and agencies are not immune. The same Q&A says there is no blanket ban on AI avatars, though using one could still be deceptive.
- Review requests. A coupon for telling people how much you loved a visit implies the review has to be positive, which the FTC treats as a problem. Check any AI-written review email for that wording.
- Influencers and affiliates. The FTC's guide says your company is responsible for what others do on your behalf, that agencies can be liable too, and that "#ambassador" is ambiguous. It expects reasonable programs to train and monitor your network. See also how affiliate marketing disclosure works.
- Claims about your own AI. If you market an "AI-powered" product, you need evidence. DoNotPay agreed to pay $193,000 over its "robot lawyer" claims.
- Copyright. The US Copyright Office concluded in 2025 that AI output is protected only where a human determined sufficient expressive elements, and that prompts alone don't count. Human work in a larger piece stays protectable. A campaign image straight from a generator may give you little to enforce if someone copies it, so build in real human selection, arrangement or editing for assets you need to own.
Where | What happens | What you do |
|---|---|---|
Meta | An "AI info" label applies to ads made or significantly edited with Meta's own tools. It sits next to "Sponsored" when those tools add a photorealistic human. Meta said in June 2026 it would start detecting third-party tools | Expect the label, and don't strip watermarks or authenticating metadata from Meta's output |
Google Ads | A label setting adds a disclosure under "How this ad was made." Campaigns targeting the EU, India or New York also get visible overlays | Label AI assets. Google says the setting doesn't guarantee legal compliance |
TikTok | Undisclosed AI content can get an ad rejected or restricted. Minor edits such as lighting or background removal don't count | Use TikTok's AIGC label or your own clear disclaimer |
New York law | Ads must conspicuously disclose a synthetic performer when the creator knows one is in the ad. Penalties are $1,000, then $5,000. Audio ads and translation-only AI are exempt. The governor's office said in June 2026 it is in effect | Log which assets use generated people, so the person setting up the ad knows |
Merlin's place in a marketing team's week
Merlin helps with the people side of marketing: the email and calendar trail between you, your agency, your freelancers and your approvers. This Mac assistant works through your Google account, mail and calendar included, so a company on Microsoft mail isn't a fit today. It won't manage ad accounts, generate creative, schedule social posts or build attribution reports, and the tools listed earlier still do that work.
- Stalled threads. The People panel shows who has been waiting on your reply, which helps when a designer's question has sat in your inbox since Tuesday.
- Prep before the call. A brief arrives shortly before each meeting with context on who's attending, such as the agency's account lead ahead of your monthly review.
- Calls with your agency. Recording working calls on your Mac is free in Merlin, and speaker-labeled transcription is a paid feature, from $29 monthly. Other recording options are in this note taker comparison.
- Nudges you sign off. In chat, request a note to the stakeholder who hasn't approved. Merlin puts that email on a card in the chat. You read and edit it, and the email waits for your confirmation.
- A weekly chase list. A scheduled routine can list what's still outstanding from your mail each Friday. Routines run unattended, so if you allow one to post to Slack, it posts without a card.
Five small marketing tests to run this week
Start with one tedious task that has an obvious right answer, and leave strategy for later.
- Audit last month's UTMs. Ask an assistant to list inconsistent values in your tagged links, then write a one-page naming rule.
- Repurpose one recording. Make clips and a recap email from your last webinar, and time the edit against how you do it now.
- Set your AI rule for suppliers. Add the clause above to the next brief and statement of work.
- Add one line to your approval checklist. Is any review, testimonial, person or result in the asset generated, and does each claim have evidence behind it?
- Start an AI asset log. Record the tool, the human edits and the platform label for each generated asset, for New York disclosures and copyright questions.
Frequently asked questions
Will AI replace marketers?
AI can take over a lot of production work, such as first drafts, resizes and clip selection. Positioning, budget decisions and accountability for claims still sit with a person, and the FTC holds the advertiser responsible either way.
Do I have to disclose that marketing content was made with AI?
No single US rule covers all of it. Disclosure depends on the case: New York requires it for synthetic performers in ads, and TikTok requires a label on significantly AI-modified ads. Separately, the FTC's rule bans AI-generated fake reviews outright, so a label doesn't fix them.
Can I copyright an ad made with AI?
Only the parts where a human determined the expressive elements, according to the US Copyright Office. Prompts alone don't qualify, but human edits and arrangements can.
Is there a free way to learn AI for marketing?
HubSpot Academy's free course runs 6 lessons, about 2 hours 49 minutes, and covers prompts, campaigns and ethics. You'll learn more by pairing it with one real task from your own calendar.
Keep your agency threads moving
Merlin shows which agency, freelancer and approver emails still need your reply, briefs you before review calls, and drafts the nudge for you to confirm. Available for Mac, working from Gmail and Google Calendar.
Get Merlin