Guide · Choosing an agency
How to choose an AI UGC agency: 9 questions and 5 red flags.
Anyone with a $300 subscription can call themselves an AI UGC agency now. This is how to tell the ones that will move your numbers from the ones reselling a tool.
Start here
Agency or tool?
Be honest about this first. It decides everything else.
- A tool is enough if someone on your team can run discovery, write hooks, brief variations and iterate on data every week.
- An agency earns its fee on the parts around the video: discovery, testing structure and iteration.
- Ask the nine questions below. Walk away on any of the five red flags.
- A good first month: discovery, a multi-angle first round, launch, and a second round built on early data.
The AI UGC tools are good and getting better. If you have a person who can learn your customer, write twenty hooks, brief variations across formats, read the ad account every week and turn winners into the next round, a tool plus that person will beat most agencies. The video is the easy part now.
An agency makes sense when nobody on the team has that time, when your first attempts came out looking like an avatar reading a script, or when the ad account is telling you the creative is the bottleneck and you need volume with judgment behind it. What you are buying is not videos. It is discovery, testing structure and iteration, with the videos as the output.
The questions
Nine questions to ask any AI UGC agency.
Good agencies answer these in a sentence each. Evasive answers are answers too.
- 01
What happens before the first ad?
You want to hear "discovery": your products, audience, messaging, competitors, KPIs, and the claims you can and cannot make. If the answer is "send us the product link", the ads will be generic.
- 02
Can I see ads with the product being handled?
Avatar-only demos are easy. A product stirred into coffee, unboxed, scrolled through on a store, compared on a bench, is the work. Ask for examples in your category or a neighbouring one.
- 03
How many variations are in a round, and across which formats?
Testing needs volume: several hooks per angle, several angles, more than one format. A round of two videos is a shoot, not a test.
- 04
Who writes the hooks, and from what?
The first two seconds decide the ad. Hooks should come from the discovery (reviews, objections, the language customers already use), not from a template list.
- 05
How do you iterate once the campaign is live?
The answer should name what they read (hook rate, hold rate, cost per result), how often, and what changes as a result. "We will make more videos" is not iteration.
- 06
How do you handle platform rules?
Meta and TikTok have rules for AI-generated content labeling and for claims, especially in health, beauty and finance. An agency that has not thought about this will cost you rejected ads or a flagged account.
- 07
Who owns the ads and the presenters?
Get usage rights in writing: where you can run the ads, for how long, and whether the same presenter can appear for a competitor.
- 08
Do you run the media, or do we?
Either answer is fine. Ambiguity is not. Know who launches, who owns the ad account, and who reports.
- 09
How is pricing structured?
Per video, per round, or a retainer. Per-video pricing suits a first test; a retainer suits an ongoing iteration loop. Ask what happens when a round does not produce a winner.
Red flags
Five reasons to walk away.
Any one of these on its own is enough.
- Guaranteed results. Nobody controls your product, your offer, your landing page or the auction. Guaranteed ROAS is a sales tactic, not a service.
- No discovery. If the process starts at "upload your product page", the ads will look like everyone else's.
- Only avatar demos. Talking heads on plain backgrounds with no product in hand means they have not solved the hard part of the format.
- No examples in your category or near it. Supplements, jewelry and software are different jobs. Ask how the angles change.
- Cannot explain iteration. If nobody can tell you what they read in the ad account and what they change because of it, you are buying videos, not performance.
What good looks like
A first month with a real agency.
Use this as the yardstick in the sales conversation.
Week one: discovery
Products, audience, competitors, reviews, objections, KPIs and the claims you can make. You should be asked questions you have not been asked by an agency before.
Week two: the first round
Several angles, each in several hooks, across at least two formats. You review for accuracy and voice, not for whether you personally like each one; the audience decides that.
Week three: live
Campaigns launch (yours or theirs, as agreed). Early signals arrive within days: which hooks hold attention, which angles get clicks, which formats convert.
Week four: the second round
Built on the data, not on opinion. Winning hooks get new bodies, winning angles get new hooks, losers are retired. From here the loop repeats, and the ads get better every round.
Where we fit
This is how Spoty Digital works: AI brand discovery, creative generation, and iteration on your live campaign data. You launch the campaigns; we make the creative and keep improving it. If that is the shape you want, book a free discovery call. If a tool and a good in-house operator is the better fit for you, we will say so on the call.
Do I need an AI UGC agency or just a tool?
A tool is enough if someone on your team can do brand discovery, write hooks, brief variations, read the campaign data and iterate every week. An agency is worth it when nobody has that time, or when your first attempts looked like an avatar reading a script.
What should an AI UGC agency show me before I sign?
Real ads, ideally in your category, with the product being handled on camera; a description of their discovery process; how many variations a round includes; and how they iterate on data. If they cannot show ads, or only show avatar demos, keep looking.
What does a good first month look like?
Discovery in the first week, a first round of ads across several angles and formats in the second, campaigns live by the third, and a second round built on early data by the fourth. Anyone promising winning ads on day one is skipping the part that makes them win.
Next step
Ask us the nine questions.
A free 30-minute discovery call with Josh. Bring the list; we will answer every one.