Should you buy technical SEO, content, digital PR or “AI optimisation”?

Commission the intervention that addresses an observed defect. Google’s current guidance says foundational SEO remains relevant to its generative search experiences; it does not prescribe a separate set of mandatory GEO files or markup. This is Google-specific guidance, not proof that all engines behave identically. (developers.google.com)

The following table is our diagnostic purchasing framework. The proposed triggers should be investigated; they are not established causal explanations of an omission.

When should technical work come before content production?

When you can demonstrate that systems cannot access the information you already have. OpenAI distinguishes OAI-SearchBot, used for search discovery, from GPTBot, used for potential model-training collection; the controls are independent. Allowing search access does not require treating training access as the same decision. (developers.openai.com)

Request tests of actual access and delivery, not merely a crawler checklist. A change to robots.txt is an implemented change; a subsequent recommendation is a separate outcome to observe.

Does a special schema package or llms.txt solve AI visibility?

Not on the evidence required to justify a guaranteed-outcome claim. Google states that its generative search features do not require special schema, and its newer guidance says Google Search ignores llms.txt for visibility and ranking purposes. (developers.google.com)

Our recommendation is to approve structured-data work for a defined technical purpose and content clarity, not because a proposal labels it “AI schema”. Require an engine-specific explanation for any additional file or delivery mechanism.

Does digital PR improve AI search visibility?

Treat it as an intervention hypothesis, not a predictable citation purchase. A useful proposal should identify relevant publishers, explain why the business has something worth covering, and separate editorial outcomes from subsequent assistant observations.

Reject fabricated reviews and paid endorsements disguised as independent recommendations. The FTC’s reviews rule addresses fake or false reviews and prohibits incentives conditioned on positive or negative review sentiment; its endorsement guidance addresses disclosure of material connections. Legitimate incentives and endorsements require more nuanced assessment than a blanket prohibition on all paid activity. (ftc.gov)

How should you assess an AI-specific website delivery layer?

Scrunch describes AXP as serving a prepared, AI-oriented mirror to selected retrieval agents while human visitors receive the existing site. That is a documented product architecture, not independent evidence of visibility improvement. (helpcenter.scrunchai.com)

Our recommended review should require:

• A comparison of human and agent versions for factual consistency.
• Named owners for product claims and updates.
• A record of which agents receive which version.
• Tests showing that citations resolve to a useful human page.
• A rollback plan and a limited initial deployment.
• Review against relevant platform policies.

Do not assume that every alternate rendering is prohibited cloaking—or that every implementation is safe. Google defines cloaking by differing content presented with the intent to manipulate search rankings and mislead users. Assess the actual design and content rather than relying on the supplier’s label. (developers.google.com)