AI didn’t replace your website. It changed what it’s for.

AI assistants now sit at the start of the buying journey. Buyers ask them what to buy, how products compare and what to look for. For the brands behind those products, that raises an uncomfortable question: if the research happens somewhere else, what is the website for?

We asked 2,000 synthetic respondents, AI-generated personas grounded in public data that model AI-assisted buyers, how they use AI and brand websites across the whole buying journey. The study, run with Evidenza, covers 1,000 B2B and 1,000 B2C respondents across the US, Europe and APAC. Three findings stood out: AI is added to the journey, not swapped in; the website remains the source of truth; and personalization helps when it’s about context, not identity.

AI use climbs through discovery, peaks in research (75% of buyers use it) and comparison (68%), then falls to 40% at the final decision and 26% at the purchase. Half of AI’s influence on the decision sits in those first two stages, research and comparison.

It also repeats your facts. Buyers use AI for technical specs (51%), prices and availability (42%) and to validate decisions (44%). If an AI answer carries an out-of-date spec or price, it travels with the buyer. Discovery is widening too: 75% say AI has increased the new brands they discover. And AI widens the funnel without eroding loyalty, since net change in loyalty to known brands is +4 for B2B and −3 for B2C.

Almost nobody stops at the AI answer: fewer than 2% do. The most common next stop is the brand’s own website (22%), level with search engines. When AI and the website disagree, 54% rely on the website and 18% on AI.

Buyers go for authoritative information (21%), detailed specs (18%), pricing (14%) and to verify something they learned elsewhere (12%). They arrive mid-conversation, and a page that can’t confirm what they were told sends them back to an AI tool.

Buyers stack sources, they don’t swap them

Heavy AI users aren’t giving up other channels. They add to them. Heavy B2B users index at 164 on analyst and trade press and 146 on newsletters. Heavy B2C users index at 134 on video platforms and 131 on marketplaces. Each of those channels carries your specs, prices and claims, so a consistent source of facts matters more, not less. The temptation is to fund an AI strategy by trimming other channels. The data says the most AI-engaged buyers are the ones using them most.

Buyers welcome personalization that helps them in the moment: answering questions in real time (+61 net helpful), tailoring to stated needs (+61) and adapting to the journey stage (+49). They push back on personalization that relies on who they are: name greetings score −12, and following up on what they viewed scores −17 (−21 for consumers). Heavy AI users are the most receptive to tailored help, which makes the hand-off from an AI answer to your site the moment that matters most.

At the purchase, 58% of buyers use the brand’s website and 26% use AI. B2C buyers are more likely than B2B buyers to name the site as the stage that most influences their final decision (21% vs. 14% for B2B). And 60% would be uncomfortable with AI completing a purchase automatically. Buyers want help deciding, not a decision made for them.

The research points to five decisions:

At Magnolia DXP, we’ve built our platform around these decisions: Magnolia Answers by ai12z for AI visibility reporting and an on-site assistant, Magnolia Agentic AI & GEO Tools for structured markup, Intent-Based Personalization for context over identity, and Magnolia Experiments by Kameleoon for the decision and checkout steps.

About the research: Evidenza surveyed 2,000 synthetic respondents across the US, Europe and APAC on behalf of Magnolia DXP. The respondents are synthetic personas generated by Evidenza and grounded in public data, not real people, so the findings are modelled behavior rather than a poll of named individuals.