Telling Clients You Use AI Before They Ask

Every agency running AI-assisted delivery right now is making a choice, whether they realize it or not: disclose proactively, or wait and hope it never comes up. Only one of those is actually a strategy. Adobe’s 2026 AI and Digital Trends report, based on a global survey of 3,000 executives and practitioners alongside 4,000 customers conducted with Oxford Economics, found that a third of customers stop engaging with a brand once they discover its content was AI-generated and they were not told.

Silence used to be the safe default, and it no longer is

For a while, the safest move for an agency using AI in its delivery was to say nothing and let the work speak for itself. That calculation has shifted. Adobe’s research found 37 percent of customers disengage specifically when they realize they expected a human and got AI instead, which is a sharper version of the same problem: it is not AI itself that erodes trust, it is the mismatch between what a client assumed and what actually happened.

The risk was never that clients would object to AI. The risk is a client finding out on their own, after assuming otherwise.

That shift shows up on the agency side too. Nearly half of agencies, 45 percent according to AgencyAnalytics’ 2026 Marketing Agency Benchmarks Report, drawn from 494 agency professionals, are now being more open with clients about how AI shows up in their workflow. Fifty-eight percent report increasing human review of AI-generated work, and 53 percent say they have raised their quality bar specifically to avoid generic-sounding output. The direction of the industry is toward disclosure, not away from it.

What clients are actually worried about, underneath the AI question

Nobody who asks “are you using AI” is really asking about the technology. They are asking three things at once: is the quality still going to be good, is anything confidential going into a prompt somewhere, and am I paying human-level rates for a process that got cheaper on your end without me knowing.

Answer those three concerns directly and the AI question mostly dissolves. Dodge them, or answer with vague reassurance, and the client fills in the worst version of all three themselves.

The disclosure paragraph, ready to use

Here is language that works, adjusted to your own voice: “AI helps draft the first version of routine posts, which speeds up how quickly we can turn ideas into content. Every single post is reviewed and approved by a person on my team before it goes anywhere near your account. Nothing publishes without that review.” Three sentences. It names the tool, states the boundary, and does not apologize for either.

What makes that paragraph work is the order. Tool first, boundary second, both stated as fact rather than as a defense. A version that leads with an apology, something like “I know some agencies overuse AI, but I promise I don’t,” invites exactly the skepticism it is trying to avoid.

Timing: the proposal stage, not the moment it gets noticed

Disclosure at the proposal stage is a feature you are describing. Disclosure after a client notices something feels off is a confession you are making. Same information, opposite framing, and the framing is entirely a function of timing.

Put the paragraph above into your proposal template or your onboarding conversation, not into a defensive email six months into the relationship. A client who hears it on day one treats it as part of how you work. A client who has to ask for it later treats it as something you were hoping to avoid mentioning, which is a much harder impression to undo than the disclosure itself would have been.

The value reframe: what changed is cost, not what the client is paying for

The instinct when a client asks about AI is to explain the technology. The better move is to name what has not changed: a human still decides what goes out under the client’s name, every time, no exceptions. That is the actual service being sold, and it did not get automated.

AI changed how fast a first draft gets produced. It did not change who is accountable for what actually ships, and that accountability, not the typing, is what a client’s retainer is paying for.

Handling the discount request without giving away margin

The discount request usually arrives dressed as a fair question: “if AI is doing some of the work, shouldn’t this cost less?” Answer the framing, not just the number. The work a client is paying for was never the mechanical act of writing a caption. It was the judgment on brand voice, the approval step that catches a bad post before it goes live, and the strategy underneath the calendar. None of that got automated.

If a concession has to happen, concede scope before you concede rate. Adding a deliverable at the existing price protects the number that actually matters to your margin far better than lowering the number itself does, and it still gives the client something tangible in exchange for asking. AgencyAnalytics’ data on rising quality bars backs this up indirectly: agencies raising their bar on AI-assisted work are not the ones racing each other to the bottom on price.

What it looks like when an agency waits too long

The failure pattern is consistent enough to describe generically, even without naming a specific incident: a client stumbles onto AI involvement themselves, usually because a post reads slightly generic, or because they saw the same phrasing pattern across a competitor’s account managed by the same agency. The conversation that follows is defensive by default, because the agency is now explaining something rather than describing it. Trust rebuilds slowly from that position, if it rebuilds at all.

Compare that to a client who heard the disclosure paragraph in the original proposal, nodded, and moved on because it was framed as part of the process rather than something to be caught doing. The difference between those two conversations is not the underlying fact. Both agencies use AI. One decided when the client would find out.

A short script for the verbal version of this conversation

The written paragraph works well in a proposal. In a live conversation, whether that is a discovery call or a check-in six months into the relationship, the same idea needs to come out more naturally. Something close to: “I want to be upfront about how we work. AI helps my team move faster on first drafts, but nothing goes out without a person reviewing it first. That review step is not a formality, it is where we catch tone problems, factual errors, anything that would not sound like your brand.” Pause after that, and let the client respond rather than filling the silence with more justification.

If the client pushes back with a specific concern, quality, confidentiality, or cost, answer that specific concern rather than repeating the general reassurance. A client worried about confidentiality wants to hear what does and does not go into a prompt. A client worried about quality wants to hear about the review step, not the tool. Matching the answer to the actual worry is what separates a disclosure conversation that lands from one that just restates the same paragraph twice.

Putting it in writing, not just in conversation

A verbal disclosure is easy to forget or misremember months later. Put a version of the same paragraph into the proposal or the statement of work itself, even briefly. One or two sentences under a “how we work” heading is enough: it does not need its own dedicated policy document to do its job here, though a fuller written policy is worth having once you are running this across more than a handful of clients.

Having it in writing also protects the agency side of the relationship. If a client’s expectations drift over time, the original written disclosure is the reference point both sides can return to, rather than relying on someone’s memory of a conversation from months earlier.

What disclosure does not change

Telling a client AI helps with drafting does not open the door to letting AI handle everything. Strategy conversations stay human. Any reply to a comment or a direct message stays human. Anything that could become a public relations problem stays fully human, reviewed by someone who can read the room in real time, not the next morning’s queue.

Disclosure is honest about what AI touches. It is not permission to expand what AI touches just because the client already knows and has not objected. Keep the boundary the same regardless of how comfortable the conversation went.

Say it before they ask, say it plainly, and keep the boundary exactly where it was before the conversation started. That is the whole playbook, and it costs nothing except the ten minutes it takes to write the paragraph into your next proposal.