Discussion starter: an agency question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- Agency
How much of your delivery now uses AI tools, and do you tell clients? Where is the disclosure line for you?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
AI now turns up in almost every agency workstream: research, clustering, first drafts, data analysis, documentation and reporting. The disclosure line I would draw has three parts. Tell clients up front, in the engagement terms, which kinds of AI tools you use and how their data is handled. Flag specific deliverables when AI produced a material part of them, such as drafted content or generated analysis. And never present AI output as expert human judgement unless a qualified person has checked it and stands behind it. What does not need disclosing is the equivalent of a spellchecker: routine assistance that does not change who is responsible for the substance.
Why disclosure is a data question before it is an ethics question
The most urgent disclosure issue is not whether a blog post was drafted by a model. It is whether client data went into a third-party AI service. Search Console exports, analytics, CRM extracts, customer reviews and internal documents can include personal data and commercially sensitive information. Under data protection law in many jurisdictions, and under most client contracts, sending that data to a new processor requires a legal basis, an agreement and often the client’s approval. Before anything else, agencies should know which tools are used, whether they retain or train on inputs, where data is processed, and whether that is covered by the client agreement.
A disclosure framework
Use of AI Disclose? How Spelling, grammar, summarising your own notes No Covered by general terms Keyword clustering, data analysis, crawl interpretation Yes, in general terms Engagement terms list tool categories and data handling Drafting client-facing content that will be published Yes, per deliverable Note on the brief or deliverable that AI assisted drafting, with named human editor Processing client data with a third-party AI service Yes, before use Client approval and a data processing agreement where required Automated reports or recommendations sent without human review Yes, clearly Label as automated; better still, review before sending What clients actually worry about
- Paying human rates for machine work. This is really a pricing question, but clients raise it as a disclosure question.
- Quality and accuracy. Whether anyone checked the output.
- Their own policies. Some organisations restrict AI use for legal, regulatory or brand reasons, and expect suppliers to follow those restrictions.
- Search risk. Whether AI-produced content could hurt their rankings. Google’s position is that AI use is not itself a problem; low-value content at scale is.
- Ownership. Whether they own the output and whether it could infringe someone else’s rights.
Addressing these directly, in writing, is more reassuring than any statement about percentages of AI use.
A policy statement that works
Hypothetical example of engagement wording:
“We use AI tools to support research, analysis and drafting. All deliverables are reviewed and approved by a named member of our team, who is accountable for their accuracy. We do not enter your personal or confidential data into AI services that retain or train on inputs, and we use only services covered by our data processing arrangements. Where published content is drafted with AI assistance, we will tell you, and it will be edited and fact-checked before delivery. If your organisation has an AI policy, we will follow it.”
Regulatory points worth knowing
In the EU, Article 50 of the AI Act, applicable from August 2026, requires disclosure when AI-generated text is published to inform the public on matters of public interest, with an exemption where the content has undergone human review and a person holds editorial responsibility. It will not apply to most marketing content, but agencies producing news-like or public-interest content for clients should understand it. Consumer protection rules on misleading content apply regardless of how content was produced.
Watch for this:
Invented expertise is where agencies get into real trouble: AI-drafted articles published under a client expert’s name without that expert reviewing them, or reports that present generated analysis as a senior strategist’s assessment. It misleads readers, breaches the expert’s trust and, on YMYL topics, can cause harm. If a named person’s authority is attached to a deliverable, that person must have reviewed it.
Disclosure and pricing are the same conversation
Clients who ask “do you use AI?” are often really asking “am I paying for time you did not spend?” Answering the disclosure question without addressing pricing leaves the real concern unresolved. Agencies that price by hours face this most sharply, because efficiency gains show up as fewer billable hours or as hours that no longer match the work. Moving towards pricing by deliverable, outcome or retained expertise makes disclosure easier: the client pays for an audit, a content programme or strategic judgement, and how efficiently you produce it is your business, provided quality and accountability hold. Whatever model you use, be ready to explain what the client is paying for and why AI does not reduce its value.
How the line may move
As AI becomes standard in every tool, disclosure of routine use will matter less, and disclosure of accountability will matter more. Clients will increasingly assume AI is used; what they will want to know is who checked the work, how their data was handled and whether the result is accurate. Agencies that set those standards now, and write them into their terms, will find the conversation easier than agencies that treat AI use as something to avoid mentioning.
Disclose data handling before any client data touches an AI service, disclose AI’s role in deliverables that will be published or relied on, keep a named human accountable for everything, and do not let AI-produced work carry a human expert’s authority without that expert’s review.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.