Is your marketing agency using AI? (And how to tell if it is helping or hurting)

By Jimmy Dodgson, Client Services Director, Maitland

Published 2026-08-12

Worried you are paying agency rates for AI output? The honest answer on where AI helps, where it must never be trusted, and how to spot slop in your marketing.

Executive summary

Your agency almost certainly uses AI somewhere. The problem is not the technology, it is the silence around it. AI genuinely helps with research, data patterns and first drafts, which frees budget for thinking. It must never be trusted with strategy, numbers, claims or final sign off. Run the swap test on your last piece of content: replace your name with a competitor’s and see if it still reads fine. Ask your agency five direct questions about where AI is used, who checks it and what it is never allowed to do. The tool was never the test. Accountability was.

Is your marketing agency using AI to produce your work? And if it is, are you getting better marketing for your money, or are you paying agency rates for output anybody could have generated in 30 seconds?

You are in the right place, and you are going to get the honest answer, including ours. I run client services at Maitland and I have spent the past few months building an agentic team of our own, so I can tell you exactly where this technology earns its place and exactly where letting it near your marketing will cost you.

Here is what this article covers: the honest state of the industry, where AI genuinely improves agency work, where it must never be trusted alone, how to spot AI slop in your own marketing in about two minutes, the questions to put to your agency, and precisely what we do and do not let it touch.

The honest state of the industry

Let us start with the answer nobody in my industry volunteers: yes, your agency is almost certainly using AI somewhere.

Research, competitor analysis, first drafts, data summaries, image work, transcript processing, code. Some of it is disclosed. A lot of it is not. Agencies have worked out that clients might feel short-changed, so the safest commercial move has been to stay quiet and let you assume everything is hand-crafted by a person in a nice office.

We think that silence is the actual problem, not the technology. Using AI badly and hiding it is indefensible. Using it well and explaining it is just good practice, in the same way that no agency hand-codes a website from scratch or manually counts keyword rankings any more. Tools change. Accountability for the output does not.

So the useful question is not "does my agency use AI?" It is "does my agency's use of AI make my marketing better or worse?" Those have very different answers, and you can find out which one applies to you.

Where AI genuinely helps

Being fair about the upside matters, because a blanket "we never touch it" claim in 2026 is either untrue or a confession of inefficiency.

Used properly, AI is exceptional at scale and pattern. It can read every competitor page in your sector in minutes. It can pull structure out of a year of messy analytics. It can process customer reviews and surface the phrases your buyers actually use, which is worth more than any brainstorm. It can produce a first draft that saves an hour of blank-page staring, and it can handle the repetitive administrative work that used to eat into the time you were paying for.

The honest client benefit is this: when the mechanical work gets faster, more of your money goes into thinking, strategy and judgement. That is a genuine improvement, and any agency delivering it should be happy to tell you so.

Where it must never be trusted alone

Now the other half, and this is where the damage happens.

AI does not know your business. It has never stood on your shop floor, met your customers, or sat in the room when you explained what actually keeps you awake. It produces the average of everything it has read, which is precisely the opposite of what marketing needs to do, because average content is invisible content.

It also cannot be trusted with the things that carry risk. Numbers, claims and statistics need human verification every time, because a confident-sounding figure that turns out to be invented is a reputational problem with your name on it, not the agency's. Strategy needs judgement about your market and your constraints. Anything that involves a promise to a customer needs a person who is accountable for keeping it.

And it should never have the final say. Everything that reaches a client of ours passes a human who is willing to put their name to it. That is not a technology policy. It is what accountability means, and it is exactly the same standard we apply to showing our working on results .

How to spot AI slop in your own marketing

Here is a test you can run this afternoon on whatever your agency last produced for you.

The swap test. Take one of your recent blog posts or social posts and mentally replace your company name with a competitor's. If it still reads perfectly well, that content is not about your business. It is a generic essay with your logo on it, and it will not move anybody towards buying from you rather than from them.

Beyond that, look for the tells. Content with no specific numbers, no named examples and no first-hand experience anywhere in it. Prose that says a great deal without telling you anything, in a register that hovers permanently between confident and empty. Certain phrases that have become the fingerprints of unedited machine output: openers about our rapidly changing landscape, endless variations on unlocking potential, and words like seamless and robust doing the work that a real detail should be doing. Copy that never once mentions a mistake, a limitation or a trade-off, because genuine expertise always includes those and generated averages almost never do.

The deeper problem is not that it sounds artificial. It is that it says nothing only your business could say, which means it earns you no trust with humans and no citations from the search engines and AI tools your buyers increasingly ask instead of searching.

The questions to put to your agency

Ask these directly. The answers matter more than the policy.

Where do you use AI in producing my work, and where do you not? Who checks the output before it reaches me, and what are they checking for? How do you verify any statistic or claim that appears in my content? What information does your AI actually work from, beyond a general prompt? And what would you never let it do?

An agency doing this well answers specifically and comfortably, because it has thought hard about the boundaries. A vague answer in either direction is the warning: agencies claiming to use no AI at all are usually either behind or not being straight with you, and agencies that cannot tell you where the human checks happen have probably not put any in.

One further point worth raising, because it is the one clients most often forget to ask about: what happens to your information. Anything shared with a general-purpose tool may sit outside your control, so your agency should be able to say plainly which systems hold your data, whether anything commercially sensitive ever leaves a controlled environment, and who inside their business can see it. If that question produces an uncomfortable pause, the policy probably has not been written down at all.

What we actually do at Maitland

We would rather show our working than make claims, so here is our setup, in the same detail I would want if I were the client.

I run an agentic team of eleven specialist assistants supporting our work, and building it taught me the single most important thing about this technology: an AI assistant is only as good as the information it has. Without proper grounding, an SEO assistant is just a general-purpose chatbot wearing a hat, producing the same averaged advice anybody else could get for free.

So we invested in what sits underneath. Each specialist works from real, owned source material rather than general knowledge. Our content assistant, for example, is built on the Endless Customers methodology by Marcus Sheridan, the same framework that drives everything we publish, which means its instinct is radical transparency and answering the questions buyers actually ask, rather than producing whatever ranks. That grounding is the difference between consistent, useful output and confident noise.

We also run an explicit list of forbidden patterns, the tells described above among them, so that generic phrasing gets caught before anything ships rather than after a client spots it. And every piece of work still passes a person who signs their name to it.

What it does not do is decide your strategy, invent your numbers, or replace the in-person Deep Dive month where we come and stand in your business. Those are judgement, and judgement is what you are paying us for.

The results of working this way show up where they should. Across all our clients in the last 12 months, the average was 52 per cent growth, and our approach to publishing, including the piece you are reading, is built on being genuinely useful rather than merely present. You can see the rest in our case studies .

The principle underneath all of this

Whatever tools an agency uses, one thing has not changed and will not.

You are not buying words, images or hours. You are buying results that can be measured, produced by people who are accountable for them. If AI helps a good team produce better work faster, that is to your benefit. If it lets a weak team produce more of nothing, faster, that is your loss and their margin.

The technology is not the test. Accountability is. It always was, and the criteria for choosing an agency have not fundamentally changed because a new tool arrived.

Frequently Asked Questions

Is my marketing agency using AI?

Almost certainly, somewhere: research, drafts, analysis, transcription or image work. Very few disclose it fully. The more useful question is how they use it and what human checks sit around it. Ask where they use it, where they do not, who verifies output and what they would never let it do.

Should marketing agencies use AI?

Yes, for the work it genuinely improves: large-scale research, pattern-finding in data, first drafts and repetitive tasks, which frees budget for thinking and strategy. No, for strategy, verified numbers, claims to customers or final sign-off. The distinction is whether a person remains accountable for what reaches you.

How can I tell if my content was AI generated?

Run the swap test: replace your company name with a competitor's and see if the piece still works. If it does, it says nothing only you could say. Also look for absent specifics, no named examples, no numbers, no mention of trade-offs or limitations, and stock phrasing about changing landscapes and unlocking potential.

Does AI content hurt SEO?

Search engines do not penalise AI involvement itself; they reward genuinely useful, experience-backed content and ignore generic material regardless of origin. Since AI tends to produce the average of what already exists, unedited output rarely earns rankings or citations. Original data, first-hand experience and specifics are what get cited.

What should I ask my agency about AI?

Five questions: where do you use it and where do you not, who checks output before it reaches me, how do you verify statistics and claims, what source material does your AI work from, and what would you never let it do. Specific, comfortable answers are the good sign. Vagueness in either direction is not.