How to Tell If Marketing Is Actually AI-Directed - 9xNine Insights
Guide Part of: AI-Directed, Not AI-Generated

Insights · Trust · Diagnostic

How to tell if marketing is actually AI-directed.

Every agency claims to use AI "the right way" now. The claim is easy to say and hard to verify from the outside. Here are five concrete questions — with what a real answer actually sounds like — that separate an AI-directed operation from an AI-generated one wearing a system's clothes.

Last updated August 3, 2026 8 min read Trust

The Problem

Why every agency sounds the same now.

"We use AI, but the right way" has become the default line in every pitch deck, whether or not it's true. The language is identical whether a senior person is actually reviewing the work or a tool is running unattended with a person's name on the invoice.

That's not necessarily dishonest — it's just unverifiable from where a buyer is standing. You can't see the review process from a homepage. You can't tell, from a finished post or email, whether anyone actually read it before it went out, or whether "AI-directed" describes a real workflow or just describes the vibe someone was going for in the sales call.

This is the checklist for closing that gap — a way to ask about the process instead of the pitch, on any vendor you're evaluating, including us.

The Diagnostic

Five questions, and what a real answer sounds like.

Ask these in a discovery call or a vendor review. The answers are more telling than anything on a website.

1

Who's accountable if it's wrong?

Real answer: a named person reviews it and owns the outcome, by name, not "our process" or "the team." Red flag: "the AI handles that," with nobody able to say who signs off.

2

Can they show you what got rejected?

Real answer: yes — a specific draft, headline, or idea that got killed because it wasn't right for the business, with a reason. Red flag: everything the tool produces goes out roughly as-is; there's nothing to point to that got cut.

3

Does quality hold steady when you're not checking in?

Real answer: yes, because a documented process governs the work, not a person cramming before a call. Red flag: output visibly sharpens right before a review and drifts the rest of the month.

4

Is there a documented safeguard, or just a vibe?

Real answer: they can describe the actual review step — what gets checked, by whom, before anything ships. Red flag: "we just make sure it's good," with no specifics when you ask what "good" means.

5

Would they pass their own test?

Real answer: they invite the question and answer it directly. Red flag: the question gets deflected back to a general claim about "quality" or "experience."

How We'd Answer

Where our Three Safeguards come from.

These five questions aren't hypothetical for us — they're close to the exact list we built our own review process to survive. The full breakdown lives in the guide this article builds on.

The short version: a named operator directs and reviews every piece of work personally (the accountability question), the process is documented rather than improvised (the safeguard question), and staying current on how AI marketing actually gets misused is treated as ongoing work, not a one-time setup. If you want the longer explanation, with examples, it's laid out in full as the Three Safeguards →

Common Questions

A few things people ask about using this list.

Should I actually ask my current agency these questions?
Yes — that's exactly what it's for. A vendor confident in their process will usually welcome the question. One who deflects it has told you something too.
What if they can't answer specifically?
That's useful information on its own, not necessarily a dealbreaker by itself. Vague answers to specific questions are worth weighing alongside everything else you know about the work.
Does "AI-directed" mean the AI never writes anything close to verbatim?
No. AI drafts plenty of the actual language. "AI-directed" describes who's setting the brief, catching what's wrong, and owning the result — not whether a model ever touched the sentence.
Is this list exhaustive?
No single checklist replaces judgment. Treat these as a starting point for the conversation, not a scorecard that produces a final verdict on its own.

Keep Reading

More on trust.

Guide

AI-Directed, Not AI-Generated

The full case: what "AI-first, human-directed" means, why it's safe, and why a directed AI beats a tool you have to drive yourself.

7 min readRead the full guide →

Coming Soon

Inside a real week: what the operator actually reviews

A walkthrough of what gets checked, changed, and rejected before anything ships — in practice, not in theory.

Coming soon

The point of this list isn't suspicion — it's that "trust us" shouldn't be the whole pitch. Ask the question, and judge the answer.

Get Started

Put us through the test.

Bring these five questions to a discovery call. We'll answer them directly — and show you what the review process actually looks like.

The Edge Series · Free

Find the edge that makes you the obvious choice.

A free 5-part email series. We help you name the one thing that already makes your business the obvious choice — the thing you're too close to see — and turn it into one move you can make this month. No pitch until the end. About 6 short emails over ~2 weeks. Unsubscribe anytime.

  • Name your obvious-choice edge — the thing you can't see from the inside
  • Turn it into one concrete move you can make this month
  • See why consistency and compounding beat one-off campaigns