AI service marketing with video creators: 3 checks beyond subscriber count
Pick video creators by viewers' jobs, demos to the end result and keeping to claim limits, not subscriber count.
Key points
- Compare viewers' jobs, not subscribers
- Demos to the end result on own input
- Keeps to claim limits set on one page
- One creator first; inquiries decide next
Contents
Compare viewers' jobs before subscriber counts
A&A perspective
When you promote your own AI service through video creators, compare the candidates by the work their viewers do, not by subscriber count. A creator partnership here means an agreement for the creator to use your service on camera in their videos.
A&A perspective
A new business that builds AI into its service often finds that the use is hard to get across, even after explaining it. Reaching many people does not add users if viewers cannot map the service onto their own work.
From the sources
Clay, a company that sells a tool for researching prospects with AI, has published a customer story about Lovable, an AI service that builds apps from plain-language requests. According to Clay, Lovable was creating something genuinely new, and the market had limited understanding of what such a tool could do or who it was for.
From the sources
The fastest way to show what was possible was through creators who could run the product on camera, such as builders on the video site YouTube. Clay describes them as the highest-ROI channel for product education and demand generation.
From the sources
For each creator, Lovable gathered average views, engagement rate, subscriber count, content themes and audience composition. Clay says it then scored each one on two questions: does the audience match the people who would benefit from an AI app builder, and is the content quality aligned with Lovable's brand?
A&A perspective
What Automate & Augment (A&A), the company writing this article, notices is that subscriber count was only one of the five things Lovable collected. The scoring question, too, was whether the audience overlapped with the people who would benefit. Applied to your own service, that becomes a check on whether the viewers do the same work as the people who would use it.
Hypothetical example
Say your AI service pulls line items out of PDF quotes into a spreadsheet. What you look at is whether the viewers handle quotes in their daily work. Subscriber count gives you a rough idea of how many people you will reach.
Check the creator demos to the end result on their own input
A&A perspective
Next, check that the creator can demo your AI service to the end result, using input they already have at hand. A demo of an AI service means showing on screen, start to finish, what goes in and what comes out.
A&A perspective
The end result is what the AI service finally produces. In the quote example, it is the spreadsheet of extracted line items. In the demo, you want the creator to show everything from feeding in the quote to the spreadsheet appearing, without skipping steps.
From the sources
The creators named in Clay's story were YouTube builders, developers and tech influencers. All of them, it says, could demonstrate Lovable in action.
A&A perspective
A&A thinks the key is whether the creator owns the input used in the demo. If the demo runs on data you supplied, viewers cannot judge whether the service would work for their own job. If it runs on what the creator handles every day, viewers can picture the documents on their own desk instead.
Hypothetical example
Here is a hypothetical comparison; neither creator is real. Candidate A has 120,000 subscribers and covers how to use AI in general. The viewers' jobs vary widely, and A has no quotes at hand.
Hypothetical example
Candidate B has 4,000 subscribers and shares ways to make office work in construction easier. The viewers are office staff at small builders. B has quotes they have handled and can talk about the accounting staff who receive the spreadsheet.
Hypothetical example
In this example, only B can demo to the end result. Partnering with A reaches far more people, but you cannot tell how many of them handle quotes.
A&A perspective
Here is the sheet to fill in for each candidate. It holds the 3 items, viewers' jobs, demo input and end result, and keeping to claim limits (covered in the next section), plus a decision to pass. It also works as the one page you show your manager when explaining which creator to choose.
| Item | What to write | If you cannot fill it |
|---|---|---|
| Viewers' jobs | Which documents or screens the viewers handle day to day | Do not guess. Drop the candidate |
| Demo input and end result | Input the creator can supply themselves, the end result that comes out, and who receives it | If you have to supply the input, it is an advert, not a demo |
| Keeping to claim limits | Given one page (effects you can guarantee, effects not to claim, the step where a person checks), whether they can talk through the demo within it | If you cannot write the page, fix your own explanation. Drop candidates who cannot keep to it |
| Decision to pass | Whether any of the three above is still filled in by guesswork | If you go ahead anyway, you will not know why it worked or why it missed |
Check the creator keeps to claim limits, using one page
A&A perspective
Third, check for each candidate whether they can keep to your claim limits when they talk. Claim limits are the line between what the creator may say in the video and what they may not. As the tool for this check, hand each candidate one page that sets out that line.
From the sources
As the first section showed, Lovable also scored each candidate on whether the content quality was aligned with its brand.
A&A perspective
Applied to your own service, A&A reads that question as whether a candidate can talk within what you are able to claim.
From the sources
Anthropic, a company that develops AI, has published a customer story about brand.ai, whose tool uses AI to keep a brand consistent. Its co-founder says today's brands exist across countless touchpoints and narratives that cannot be contained in a static guideline document.
From the sources
The same story says the traditional approach of static brand guidelines written by agencies no longer works when teams around the world create content constantly.
A&A perspective
The source is about running a company's brand, not about creator partnerships. Even so, A&A thinks the two share one point: a thick document does not decide the words that end up in the video. So rather than handing over a brief after you partner, check before partnering whether the candidate can talk within the limits.
A&A perspective
The page holds three things: the effects you can guarantee, the effects not to claim, and the step where a person checks. AI output can contain errors.
Hypothetical example
In the quote example, the effect you can guarantee stops at pulling line items out of a PDF quote into a spreadsheet. Saying the spreadsheet can go straight to accounting unchecked is an effect not to claim. You do ask the creator to say in the video that a person checks the extracted results.
Hypothetical example
One way to check: hand the candidate the page and ask them to talk through the demo in their own words. Keep the candidates who mention the step where a person checks, without saying the spreadsheet can go straight to accounting.
A&A perspective
If you cannot write this page, the creator is not the problem. It is a sign that you have not yet put into words who your service helps and with what. In that case, fix the description of your service before adding candidates. Record what you find in the sheet shown earlier, in the row on keeping to claim limits.
One creator first, then let the results decide
A&A perspective
Approach one creator first. After that one, look at the jobs of the people who send inquiries, and then decide whether to add candidates.
From the sources
According to Clay's story, Lovable's team ran creator partnerships by starting small, testing the hypothesis and adding volume to what worked. As one of them put it, "When something hits, then we start adding volume."
From the sources
In early 2023, Lovable's founder released a side project to help developers move faster with AI. It went viral, but Anthropic's story says the people who showed up to use it were not the developers he had built it for.
A&A perspective
A&A puts the two side by side, starting small and unexpected people showing up, to decide what to watch after the first creator. Watch the questions and inquiries that arrive after the demo rather than the view count, and check whether the jobs of those people match the viewers' jobs you assumed on the sheet.
A&A perspective
One creator cannot tell you everything. The going rate for fees, whether you can keep working together, and whether another creator would get the same result are not settled by a single result.
From the sources
On Lovable's creator partnerships, Clay's story says a small team built hundreds of partnerships in a matter of days.
A&A perspective
That figure comes from a customer story published by the company selling the tool, and no one outside has verified it. Lovable was also a company whose demand had already taken off. It is no basis for your own forecast; if your service is not yet known, expect refusals to be the norm.
Summary
A&A perspective
Compare partner candidates by their viewers' jobs before subscriber count.
A&A perspective
Check that the creator can demo your AI service to the end result on their own input.
A&A perspective
Check that each candidate keeps to the claim limits written on one page you hand them.
A&A perspective
Start with one creator. Let the jobs of the people who inquire decide the next step.
When choosing video creators, treat subscriber count as a rough guide to reach and narrow the field by viewers' jobs, by whether they can demo to the end result and by whether they keep to claim limits. What you decide next is what goes on the one page you hand candidates and which single creator to approach first.
FAQ
Should I ignore subscriber count?
No. Use it to estimate how many people you will reach, not to decide who to partner with. Decide on the viewers' jobs, on whether the creator can demo to the end result and on whether they keep to the claim limits.
What is the going rate for paying a creator?
This article has not checked it. Before partnering, you need to check the going rate in your market and the rules on disclosing that a paid video is an advertisement. A single result will not tell you a fair fee either.
Does this work for a service that takes weeks to deliver?
Not well. This way of choosing works for services that reach the end result in a short time. A service that takes weeks to deliver cannot be demoed, so A&A thinks referrals or one-to-one proposals fit better.
Sources & editorial note
Primary pages read for this article. Publication dates below belong to the sources; access dates record our research.
- How Lovable uses Clay to scale one of the fastest-growing startups in history
Clay · undated customer-story page; no visible publication date
Accessed 2026-10-24 - Lovable helps anyone create software 20x faster with Claude
Anthropic · undated case-study page; no visible publication date
Accessed 2026-10-24 - Brand.ai uses AI to make brands more human with Claude
Anthropic · undated case-study page; no visible publication date
Accessed 2026-10-24