“AI personalization” can describe several different jobs. A system may draft a headline, choose from approved alternatives, or help evaluate a test. Before comparing products, ask which decision the AI is actually making.
This is a way to evaluate vendor explanations, not a report that a new industry standard has arrived. It also avoids treating every automated action as evidence of a better visitor experience.
Ask what the AI produces
Start with a concrete output. Does the demonstration show proposed text for a person to review, or a message already placed on a live page? Those are different responsibilities.
Ask where the approved product facts come from and whether the proposed copy can introduce a new promise. Request a demonstration with a fictional example so you can discuss the workflow without supplying private customer information.
Ask what decides which message appears
A system might apply a configured rule or select a message through another process. Ask the vendor to explain the actual setup rather than inferring it from an animated dashboard.
Use two example visits with different known conditions and ask what each would see. Then ask what happens when information is missing. Record answers that still need demonstration; do not turn them into confirmed capabilities in your comparison notes.
Keep selection separate from evaluation
Showing a variation proves only that a variation appeared. It does not by itself establish why a visitor acted or whether the change caused a business result.
Ask what outcome is measured, how the comparison is made, and what would make the result uncertain. A vendor should be able to distinguish a product demonstration from evidence gathered in an actual test. You do not need to accept an unexplained “winner” badge as the whole evaluation.
Locate the human decisions
Find out who approves the copy, changes the eligible audience, and decides whether an experiment should continue. Ask how the team would investigate an unexpected message and which decisions remain with your own staff.
These questions matter even when the technology is impressive. A clear division of responsibility gives your team something to evaluate; a broad promise to “let AI handle it” leaves important details open.
Write down what was demonstrated
Use three headings in your notes: drafting, choosing and evaluating. Under each, record what you saw, what the vendor explained, and what remains unverified. Include the date and the applicable product or plan.
This can reveal that two products using the same label perform different jobs. It can also show that a tool addresses only part of the work your team needs. Neither finding requires declaring static pages obsolete or promising a conversion increase.
1976 PE's current public copy describes personalized headlines and calls to action alongside AI testing. Treat that as a reason to ask these specific questions, not proof of an outcome for your site. Watch the Demo and note which decisions are shown.