Issue #33: Use It, Don't Trust It 💎

Issue 33
May 21st, 2026

I caught myself doing something this week that I'm pretty sure a lot of you do too. I was writing a note, asked Claude to take a first pass, read it back, and then rewrote basically the whole thing anyway. Not because it was bad. It was fine. It just wasn't mine. The structure was useful, the voice was a stranger's, so I kept the structure and threw out the voice. Which is what I do every single time. Which made me realize I don't actually trust any of these tools to ship something with my name on it. I trust them to get me unstuck, and that's a very different thing.

Turns out that's not a me problem. It's the whole creator economy right now. On Tuesday, Kit (the email platform that used to be ConvertKit) published a survey of 550 working creators about how they really use AI, and two numbers do all the work. 57.3% use AI every single day. And 89.2% always review and edit the output before anything goes out. Zero percent, not a rounding error, zero, said they trust it fully.

Here's the thing. We've spent three years arguing about whether AI was going to replace creators. That argument is over, and not because one side won the debate. It's over because creators already answered it with their behavior. They use it constantly. They hand it nothing. So the real fight, the one that actually matters now, isn't adoption. It's control. Who edits, who approves, whose face, whose voice, who gets to hit undo. And the same week Kit put numbers on it, YouTube walked onto the Google I/O stage and proved the point in public.

In this issue:

Let's dig in.

The debate about whether creators would use AI is over. The fight about who's allowed to drive just started.

The Kit survey is the first piece of data I've seen that actually describes how working creators use AI instead of how a thinkpiece imagines they do. 57.3% use it daily, 71.7% at least weekly, and only 17.6% never touch it. That alone resets the conversation. Daily AI use is now the default for people who make a living creating, and it's running ahead of the broader workforce. The "is this a fad" question is settled.

What's more interesting is what they're using it for. The top two use cases tied at 82.7%: writing and editing, and brainstorming. Research and summarizing came in at 72.8%. Analyzing data at 48.2%. Read those again and notice the pattern. The heaviest usage is upstream of the deliverable, not the deliverable itself. Creators aren't mostly using AI to produce the thing. They're using it to think, to get unstuck, to react against a bad first draft so they can find the good one. When Kit asked what creators want AI to do inside their own tools, the answers stacked the same way: 63.4% want it to summarize what's working in their account, 53% want subject line ideas, 52.3% want help building automations. Only 39.3% want it to draft from scratch. People want AI on the operational grunt work they procrastinate on, and they want to keep their hands firmly on the creative.

Then the number that everything else bends around. 89.2% of creators always review and edit AI output before it goes anywhere. Only 3.5% use it with minor tweaks. Nobody, literally 0%, ships it untouched. And when Kit asked how comfortable creators would be letting AI take actions inside their account automatically, only 12% said "set it and forget it." A majority, 56%, said they'd want a preview, an undo, and an audit log of everything the AI did before they'd be comfortable. That is not the behavior of people afraid of AI. It's the behavior of people who've decided exactly how much rope to give it.

One honest caveat, because it matters. Kit is a vendor, it's selling AI features, and this sample skews toward business-minded creators (59.3% full-time, heavy on coaches, course sellers, and newsletter writers). It is not a read on the average 19-year-old posting TikToks. So treat the precise percentages as directional. But the shape of the finding is almost certainly real, because it matches what you can watch creators do everywhere: lean on the machine for leverage, refuse to let it speak for them.

Here's why this is the story and not just a stat. The same week, YouTube used Google I/O to push Gemini deeper into creation than it's ever gone, and the headline wasn't the AI. It was the controls bolted to it. Every Gemini Omni remix carries a watermark and identifying metadata, links back to the original, and creators can opt out of visual remix anytime. Likeness detection just opened to every creator 18 and up. YouTube is not asking whether creators want AI in the building. It's racing to ship the locks, the labels, and the off switch, because it has clearly figured out the same thing the Kit data shows: adoption is a settled question, and trust is the actual product now. The platform that wins the next few years isn't the one with the flashiest generation. It's the one that lets creators keep control while using it.

The bottom line: Spend an hour this week auditing your own AI workflow, and sort everything into two piles. Pile one is what you let AI do: first drafts you'll rewrite, repurposing a video into captions, summarizing your own analytics, brainstorming angles. Pile two is what you let AI decide: your point of view, the final edit, what's actually true, who you are on camera. The 89% who always edit aren't being precious. The edit is the part with their name on it. That's the whole job in 2026, honestly: get ruthless about offloading pile one so you have more hours for pile two, which is the only part nobody can do for you. The Pro-Tip below has the practical version.

Three data points that tell a bigger story.

YouTube's Google I/O drop on Tuesday was three announcements wearing a trench coat, and together they tell you exactly where this is going. Ask YouTube, the conversational search feature we flagged as a test back in Issue #31, is graduating toward a full rollout: you ask a real question, it compiles videos from across the catalogue and lets you ask follow-ups. Gemini Omni came to Shorts Remix and the Create app, so anyone can retheme a scene or drop themselves alongside a creator with a text prompt. And the quiet one that matters most: likeness detection, the deepfake-spotting tool that only reached celebrities and their agencies a few weeks ago (we covered that in Issue #30), is now available to all creators 18 and older. Notice the design. Every remix is watermarked, links back to the source, and is opt-out-able. That's not a coincidence, it's the strategy. If you're on YouTube, the action item is unglamorous and real. Go enroll in likeness detection now that you can, and make a deliberate choice about Shorts remix instead of letting the default choose for you.

Two announcements this week, same machinery. At TikTok World, TikTok rolled Creator AI Search into TikTok One, a tool that reads a brand's campaign brief and a creator's profile and recommends who to hire. The same week, Marketing Week reported X is launching Creator Connect, an AI matcher pairing brands with creators by objective, audience, and real-time trends. AI matching is moving from a feature you opt into to the default front door of brand discovery. Which means a model you don't control is increasingly deciding whether you even get surfaced for a deal. Here's the part you do control. These systems read your profile, your niche signals, your category clarity, and your proof of past work. The creators who get matched are the ones whose positioning is legible to a machine: clear lane, obvious category, receipts attached. Vague bios and "I make a bit of everything" accounts are about to get quietly skipped. Treat your profile like an input to an algorithm that's pitching you to brands while you sleep, because that's now literally what it is.

A fresh Upfluence study of 17,288 creator applications found brand-deal demand rotated hard in the first quarter. Fashion/Apparel gained 6.6 points of application share and Home/Living gained 6.2, with CPMs up roughly 50% in both. Meanwhile Health/Wellness gave back 6.3 points, with Sports/Fitness and Food/Nutrition also sliding, and CPMs in those categories down anywhere from 30% to 55%. Entertainment and Events posted the wildest swing, a reported 268% jump in median CPM. The usual caveat applies: this is Upfluence's own applicant pool, so it reflects who's pitching through their platform, not the entire economy. But the directional signal is useful and actionable. The January "new year, new me" wellness money has cooled, and spring lifestyle spend (clothes, home, going out again) is where budgets moved. If you sit anywhere near fashion, home, or experiences, this is the quarter to be pitching aggressively. If you're in wellness or fitness, it's a reminder that your pricing power is seasonal, and undercharging now because deals feel scarce will haunt you when demand swings back.

What else shipped this week.

TikTok launched TikTok GO, a US feature that lets people discover and book hotels, attractions, and tours without leaving the app, through videos, search, and location pages. Launch partners include Booking.com, Expedia, Viator, GetYourGuide, Tiqets, and Trip.com. Creators who feature these places can earn through commissions and creator campaigns.

Why this matters for creators: This is the single most important thing in the issue for travel, food, and local creators, and it's easy to underrate. For years your problem was attribution. You drove a thousand people to a hotel and had no way to prove it, so you got paid for views instead of the bookings you actually generated. TikTok GO closes that gap inside the app, the same way TikTok Shop did for products. The move right now is to start tagging the places you already feature and treat the booking data as the asset. A creator who can walk into a tourism board or a hotel group and say "my content drove 400 bookings last quarter" is no longer selling reach, they're selling revenue, and that's a completely different price tier. Get in before the lane fills up.

Meta added a visual Content Planner and bulk upload options across Instagram and Facebook, aimed at creators who batch and schedule.

Why this matters for creators: Useful if you batch, with one big asterisk. Creators in the scheduling communities are still reporting that Meta's publishing tools fail in ways that cost real money, posts that don't go out, delayed publishes, duplicates. A prettier calendar doesn't fix a scheduler you can't trust to actually publish when a campaign deadline is on the line. So by all means try the planner, but don't fully cut over from a tool that's been reliable for you until you've watched this one publish cleanly for a few weeks. Infrastructure earns trust by working, not by launching.

DAX US and Captivate announced an integrated monetization partnership launching in June, bundling programmatic ads, direct sales, feed drops, memberships, subscriptions, and tipping into one workflow for podcast creators.

Why this matters for creators: Podcast monetization has always been a mess of disconnected tools unless you were big enough to have a sales team. Bundling it is genuinely helpful for the middle of the market. The thing to watch, same as with any of these creator operating systems, is whether it actually simplifies your revenue or just adds one more marketplace taking a cut in the middle. Read the rev-share terms before you route your inventory through it. Convenience that quietly costs you 30% isn't convenience.

Draw your AI red line before a platform draws it for you.

The Kit data gives you a framework most creators are using by instinct without ever writing it down. So write it down. Take ten minutes and split your work into two explicit lists: what AI is allowed to touch, and what it never is.

The "allowed" list should be longer than your gut wants it to be, and it should be everything operational and upstream. First drafts you're going to rewrite anyway. Turning one long video into ten captions. Summarizing your own analytics so you stop avoiding them. Brainstorming twenty title options so you can throw out nineteen. Writing alt text. Drafting the boring follow-up email. This is the pile that eats your week and doesn't carry your fingerprint, and handing it over is how the 89% buy back the hours they spend on the part that does.

The "never" list is shorter and it's sacred: your actual point of view, the final edit, the call on what's true, your face and voice on camera, and the relationships with your audience and your brand partners. That's the stuff people are paying youfor. The second AI starts making those calls, you've automated away the only thing that was scarce.

Then go turn on the controls that now exist. If you're on YouTube, enroll in likeness detection now that it's open to everyone 18 and up, and make a deliberate choice on Shorts visual remix instead of letting the default decide for you. If you use AI inside any tool that can take actions on your account, find the preview and undo settings before you need them. The whole game, the thing the data and the platforms are both screaming this week, is that leverage and control aren't opposites. The creators who win the AI era are the ones who use it the most and trust it the least.

That's all for this week. Go use the robots for everything on your "allowed" list, and guard the "never" list like it's the business, because it is. And if someone forwarded this to you, sign up to get your own issue every Thursday.