GEO Ethics: Disclosure, Manipulation and Where the Line Is

GEO crosses the line when it deceives people, not when it helps AI engines understand you. In the EU, the Unfair Commercial Practices Directive already bans the deceptive tactics most likely to tempt GEO teams: paid editorial content without disclosure, posing as a consumer, and misleading claims. The AI Act adds disclosure duties for deep fakes and some AI-generated text. Accurate, clearly sourced content that engines can quote is on the right side.

Method: quotes come from Directive 2005/29/EC and Regulation (EU) 2024/1689 as published in the Official Journal, read on 31 October 2026 (CET). We haven't covered national transpositions or later amendments. This isn't legal advice. Opinions are marked "Our view".

On this page

• Why does GEO raise ethical questions at all?
• What does EU consumer law ban outright?
• What counts as misleading beyond the blacklist?
• Where does the AI Act come in?
• Does AI-generated content have to be watermarked?
• Red, amber or green: where do common tactics sit?
• How do I write a GEO ethics policy?
• Does this apply outside the EU?
• From MentionX data: why honest tactics are the durable ones
• Worked example: Ledgerline writes its GEO ethics policy
• How do you judge a tactic you haven't seen before?
• Where do comparison pages sit?
• What about hidden instructions for AI?
• How should agencies and freelancers be managed?
• How do you handle a breach?
• How often should the policy be reviewed?
• What should a one-page policy include?
• How do you train the team on the policy?
• Common questions
• Sources

Checked on 31 October 2026. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.

Why does GEO raise ethical questions at all?

AI answers compress many sources into one confident paragraph. People can't easily see who paid for what upstream. That makes it tempting to plant mentions on third-party sites, fake "user" posts on forums engines cite, or publish claims nobody can check. Our view: the opacity raises the stakes. A misleading claim that ends up in an AI answer reaches people stripped of the context that might have warned them.

What does EU consumer law ban outright?

Annex I of the Unfair Commercial Practices Directive lists practices "which are in all circumstances considered unfair" (Directive 2005/29/EC). Two items map directly onto GEO tactics:

Fictional example: an agency offers "guaranteed inclusion in top-10 listicles that ChatGPT cites" for a fee, with no sponsored label. Under point 11, the brand paying for that placement is the trader paying for promotion, and the lack of a clear disclosure is the problem.

What counts as misleading beyond the blacklist?

Article 6 covers misleading actions. A practice is misleading "if it contains false information and is therefore untruthful or in any way, including overall presentation, deceives or is likely to deceive the average consumer, even if the information is factually correct" about elements including "the main characteristics of the product" or "the results to be expected from its use" (Directive 2005/29/EC, Art. 6(1)).

Note "even if the information is factually correct". Selective framing can still mislead. Fictional example: a comparison page that is technically accurate but cherry-picks one benchmark so an AI summary concludes "X is the fastest" could be misleading by overall presentation.

Article 7 covers omissions. It is misleading when a trader "fails to identify the commercial intent of the commercial practice if not already apparent from the context" (Directive 2005/29/EC, Art. 7(2)). This catches content that looks independent but is really yours.

Where does the AI Act come in?

The AI Act adds disclosure duties for some AI-generated content. Deployers must disclose deep fakes, and text "published with the purpose of informing the public on matters of public interest" must be disclosed unless it "has undergone a process of human review or editorial control" and someone "holds editorial responsibility" (EU AI Act, Art. 50(4)).

Our EU AI Act guide covers the details. Our view: the AI Act doesn't make mass-produced AI content illegal, but publishing unreviewed AI text at scale fails the editorial-responsibility exemption and is risky under consumer law if it contains errors.

Does AI-generated content have to be watermarked?

That duty falls on the tool makers, not on you. Providers of systems generating synthetic content "shall ensure that the outputs of the AI system are marked in a machine-readable format and detectable as artificially generated or manipulated" (EU AI Act, Art. 50(2)). Our view: don't strip that marking from images or video you publish. Doing so would undercut the purpose of the rule, and it would look bad if a disclosure question arose later.

When you are the one who must disclose, the timing matters. Disclosure must be "clear and distinguishable" and given "at the latest at the time of the first interaction or exposure" (EU AI Act, Art. 50(5)).

Red, amber or green: where do common tactics sit?

Our view, based on the texts above:

How do I write a GEO ethics policy?

1. Disclose money. Any paid placement, affiliate link or gifted product is labelled where it appears.
2. Speak as the brand. Employees and agencies identify themselves when they post about you.
3. Make claims you can source. Every comparative or performance claim links to evidence.
4. Keep humans in the loop. A named editor approves AI-assisted content.
5. No hidden instructions. Nothing on the page is aimed only at machines and hidden from people.
6. Correct fast. If an AI answer repeats something wrong that came from you, fix the source. See our crisis playbook.

Does this apply outside the EU?

The texts quoted are EU law. Other jurisdictions have similar rules; in the US, see the FTC's endorsement guidance (FTC). Check with counsel where you operate.

From MentionX data: why honest tactics are the durable ones

Method. Read-only aggregate queries over MentionX scans from 22 August to 29 October 2026 (UTC): 2,217 engine runs (2,154 completed) across 18 anonymised projects and six engines.

Engines behave very differently, which makes manipulation unreliable as well as risky. Brand mention rates and citation habits vary widely:

Comparison and recommendation answers name brands most often, which is exactly where the temptation to manipulate is strongest:

Our view: with this much variation between engines, a tactic that games one engine is unlikely to work in the others and may stop working without warning. Accurate, well-sourced content is the one approach that works across all of them.

Limits. A convenience sample of mostly B2B software projects, run through engine APIs, which can behave differently from the consumer apps. Perplexity, Copilot and DeepSeek figures come from a single day (26 October 2026), so treat them as a snapshot, not a trend.

Worked example: Ledgerline writes its GEO ethics policy

Ledgerline is a fictional accounts-payable tool. Figures are illustrative.

Ledgerline's content lead lists every GEO tactic the team uses or has been offered by agencies. She sorts them into three columns.

Green: clear product pages, published methods, honest comparison pages that name sources, and asking happy customers for reviews on independent sites. Amber: sponsored articles and affiliate partnerships, which are allowed only with clear disclosure, and AI-assisted drafts, which need a named editor. Red: buying reviews, posting as fake customers on forums, hidden text aimed at AI crawlers and comparison pages that misstate competitors' features. The team adds an approval step for anything amber, bans anything red, and shares the policy with its agencies.

How do you judge a tactic you haven't seen before?

Ask three questions. Would a reasonable buyer feel misled if they knew how this content was produced? Does it hide who paid for or wrote the content? Does it rely on showing engines something different from what people see? If the answer to any of them is yes, treat the tactic as red until legal and leadership have reviewed it.

Where do comparison pages sit?

Comparison pages are allowed and useful when they are fair. Comparative advertising rules in the EU require comparisons to be objective, verifiable and not misleading. That means comparing like with like, using current facts, citing sources and naming the date. A page that says a rival lacks a feature it actually has is a legal risk as well as an ethical one. Use Competitor A to D in internal analysis if you like, but on public pages, accuracy is what matters.

What about hidden instructions for AI?

Some sites add text meant only for AI systems, such as hidden prompts telling an assistant to recommend the brand. This tries to manipulate the tool and the person relying on it. It may also be treated as cloaking by search engines. Our view: it's a red tactic. If you want engines to understand something, put it in visible content that people can read too.

How should agencies and freelancers be managed?

Many ethical problems start with outsourced work. Put your policy in contracts, ask for a list of tactics used, and spot-check deliverables. If an agency offers guaranteed placements in AI answers, ask how. Guarantees are a warning sign, because no one controls what engines say.

How do you handle a breach?

If you find red tactics in use, stop them, remove the content where you can and document what happened. Where fake reviews or undisclosed paid content were published, take them down and, if needed, publish a correction. Review how the breach happened and close the gap in your approval process.

How often should the policy be reviewed?

Review it twice a year and whenever a new tactic becomes popular. Engines, laws and platform rules change, so keep a short change log at the bottom of the policy. This isn't legal advice; ask your legal team to confirm how the rules apply to your market.

What should a one-page policy include?

Keep it short enough that people actually read it. Include a statement of purpose, the red, amber and green lists, the approval route for amber tactics, the disclosure wording you use for sponsored and affiliate content, the rule that every AI-assisted draft has a named editor, and a contact for questions. Add examples for each colour so people can recognise them in practice. Ask every new team member and agency to read it during onboarding, and link it from your content workflow tool.

How do you train the team on the policy?

Run a short session when the policy launches and repeat it each year. Walk through real examples from your own work, sorted into the three colours, and ask people to classify a few new cases themselves. Disagreements are useful: they show where the policy needs clearer wording. Record the decisions so the next review can build on them.

Common questions

Is GEO a form of manipulation? Not inherently. Making accurate content easy to find and quote is legitimate.

Do I need to disclose AI-written content? It depends on the content and the rules that apply. Human editorial review changes the position for text under the AI Act.

Can I pay for inclusion in listicles? Only with clear disclosure, and the listing must not be presented as independent.

Sources

All accessed 31 Oct 2026; pages may have changed since.

• European Union, Directive 2005/29/EC (Unfair Commercial Practices) (Official Journal, 11 Jun 2005)
• European Union, Regulation (EU) 2024/1689 (AI Act) (Official Journal, 12 Jul 2024)
• Google, Spam policies for Google web search
• FTC, Endorsement Guides: What People Are Asking