Our clients know when we use AI. We tell them whenever it contributes to something we create. That is integral to our transparency policy, not a conversation we wait for a platform to force upon us.
But telling a client how something was made and understanding how their customers will react to that information are different responsibilities.
Imagine a campaign we have developed together. The client understands the process, approves the work and is comfortable with the use of AI. Then we show how it will appear in a feed, with the platform’s AI information attached.
“Will that change what people think of us?”
That is the question. Not whether the work is good, not whether the expertise is genuine, not whether we have been open, but whether the label itself could influence how the business is judged.
An explanation does not equal acceptance
In my last article, “It’s not really yours”, I described what happened when I shared AI-augmented work with my personal Instagram audience, I labelled it and explained the creative process. People still objected. They were not telling me the images were bad. They were saying that using AI to make them was wrong.
More detail did not resolve that disagreement. My explanation addressed the thought and care behind the work. Their objection was to the method itself.
That experience does not predict how a B2B buyer will respond or isolate the effect of a platform-applied label. It does establish something I cannot ignore: not everyone who objects simply needs a better explanation.
What the platform says, and what people hear
Instagram and LinkedIn do not present AI information in identical ways. Meta’s published approach to organic posts uses industry-standard signals and creator disclosures. It distinguishes generated content from content it detects as only edited with AI, placing information about the latter in the post menu.[1]
LinkedIn’s Content Credentials feature displays an icon on images and videos carrying C2PA information. That can include camera photographs, not just AI-generated material. The icon indicates that provenance information is available; clicking it reveals the details. It is not, by itself, an “AI-generated” verdict.[2]
Those distinctions matter. We should assess the actual placement, wording and available information, rather than imagine one universal AI warning across every platform and format.
C2PA, the standard behind Content Credentials, can provide signed information about an asset’s origins, tools and editing history.[3] That can help someone understand what happened. It cannot make them approve of it.
Someone asking whether an image was generated or merely edited may find the detail useful. Someone who objects to generative AI may find that the same detail simply confirms their objection. Others may never open the information behind the icon.
This is not an argument against Content Credentials. It is an argument against expecting a technical record to resolve a disagreement about values.
When clients ask: “What will this say about us?”
For clients, the conversation begins with a process they already understand. They are not discovering that AI was used; they are considering what its visibility might mean under their own brand.
A client can approve the production method and still be concerned about the audience’s interpretation. “We are comfortable with this” is not the same as “our customers will be comfortable with this”.
Our responsibility, as I see it, extends beyond explaining which tools we used. We should bring the likely platform presentation into the creative discussion. Does the label introduce a question the campaign was never intended to raise? Will the client end up defending a production choice when it wants to discuss a customer problem?
That might be a manageable concern. It might not be a concern at all. But client approval alone cannot answer it. If an unexpected audience reaction follows, the question becomes why it was not considered together before publication.
When prospects ask: “How much of this is yours?”
A prospective client may first encounter a business through a LinkedIn case study. The project is real. The results are evidenced. The customer’s quotation is authorised. An accompanying illustration carries information identifying AI use.
Will the reader understand that the information refers to the illustration? Or will it make them question the wider account?
The same uncertainty could affect thought leadership. A director’s insight can be entirely genuine, but AI information attached to the supporting image might change how a reader credits the thinking beside it.
A proposal could then be read through that earlier impression. “Was this developed around our business, or generated from a template?” The team may have spent weeks understanding the brief. The question is whether visible AI use elsewhere makes the prospect less willing to recognise that contribution. These are possible reactions, not documented outcomes. Nor do these platform features mean that AI-written proposals are automatically detected and labelled. A label on an image is not a finding about the writing beside it, but marketers cannot assume audiences will keep that distinction, or that a label will leave their belief in the expertise behind a brand unchallenged.
When customers ask: “What does this say about the brand?”
When we publish marketing for a client, its audience sees the client’s name, not a record of the agency’s decisions. Any judgement about AI may therefore attach to that brand.
Consider a hypothetical professional-services campaign built around personal attention. The claims are accurate, the imagery is illustrative and the execution is strong. The client has approved the use of AI.
A customer might nevertheless see a contradiction: a business promising human attention while appearing to automate how it communicates. Another customer may see a sensible use of technology, while someone else objects on principle.
Confidence in the work alone cannot resolve those differences. A client can approve a production method, but it cannot approve it on its customers’ behalf.
Does disclosure affect trust?
There is evidence that disclosure can affect trust. In thirteen experiments, Oliver Schilke and Martin Reimann found that actors who disclosed AI use were trusted less than those who did not. They linked the effect to perceived legitimacy: whether people considered that use appropriate. The effect also appeared where AI involvement was already known.[4] Being exposed by a third party, however, damaged trust more than disclosing it.
That supports taking the question seriously, but it does not tell us how much a particular LinkedIn label will affect a specialist B2B campaign.
My experience is a warning against assuming acceptance, not a forecast of universal rejection. We need to allow for indifference, positive responses and negative ones, then establish which matter for the people we need to reach. A change in trust is not automatically a lost contract. Equally, a campaign can attract attention without improving someone’s opinion of the brand. Those outcomes need to be examined separately.
How to plan for the label, not just the asset
- Start by viewing the work where people will encounter it: the expected feed presentation, including the label or icon and what appears when someone opens it. Check the relevant platform and format rather than rely on a generic mock-up of an AI badge.
- Test without telling people what to notice. What did they see? What did they infer? Did their view of the business change? Only then share the fuller provenance information and explore whether it answers the concern. That separates an unclear label from an objection that survives clarification.
- If possible, use a closed research exercise to compare the same asset with different disclosure presentations, including an unlabelled control. That is a way to investigate the label’s effect, not a reason to conceal AI use in public.
- Assess audience comfort. Work with the client to put the labelled work in front of a sample of its own customers and prospects and explore whether it changes their perception of the brand and willingness to engage. Separate established customers from people interacting with the business for the first time, as familiarity can change how much weight they give the lavel.
- Be prepared to change the production route. A fuller explanation may help, but where the objection remains and the risk is unacceptable, the answer may be a different way of making the work. Removing accurate provenance or hoping nobody notices is not the solution.
Disclosure is the starting point.
Transparency around AI use remains essential. What needs to change is the assumption that disclosure completes the job.
Part of our role is to help clients assess that potential impact before it becomes their problem.
We can explain how the work was made. We cannot require the audience to approve of it. We have to plan for both.