The way buyers find and evaluate vendors has shifted faster than most marketing plans.
In September, Kinsta published a consumer survey run by Propeller Insights across 1,008 U.S. adults. 59.2% use AI tools for research often or occasionally, and 61.2% expect AI to replace websites as the main way people discover products. Kinsta sells web hosting, so read it with that in mind, and note that every figure is self-reported. The pattern still matches what we hear from buyers.
sometimes or often skip websites entirely while researching with AI
always trust an AI product recommendation without further research
visit the company's website always or most of the time after an AI recommendation
The reasons people give for skipping websites are worth reading closely. Among those who do it often, 52.2% say AI is faster and 45.2% value being able to ask follow-up questions. The rest is a complaint about the web itself: 42% point to ads and pop-ups, and 37.6% say search results feel cluttered and written for algorithms. One trade analysis also notes that the base for the 53.8% figure is not clearly stated. Treat it as directional.
Trust, then verify
Buyers use these tools constantly without believing them. 58% trust content less once they suspect AI wrote it, 52.1% still trust traditional search results over AI answers, and 90.7% are concerned about AI-generated misinformation. 86.2% believe they encounter AI content regularly, and 77.3% are confident they can identify it. We would place less weight on that last figure. Confidence in spotting a fluent, wrong answer tends to run ahead of the ability to do it.
What happens next is the part that matters for any company with something to sell. After an AI recommendation, 44.7% of buyers visit the company's website always or most of the time, and another 25% do so occasionally. 61.1% associate the quality of a company's website with the quality of its products or services. The assistant opens the conversation. Your site is where the buyer checks it.
The same control, one layer up
We apply this principle at the device edge. In an Edge Data Trust Assessment, data from connected devices is validated at the gateway, before it reaches a hyperscaler's AI models, because the model inherits whatever the edge lets through. Under the shared responsibility model, the integrity of that data stays with the customer.
A public website now plays the same role for buying decisions. It serves two audiences at once: the answer engines that read and summarize it, and the people who arrive to confirm what those engines said. Both need the same facts, current and provable. The survey adds one more wrinkle: 43.3% say knowing bots are active online makes them trust websites less. Crawler traffic is a governance question as well as a capacity question.
Five controls for the verification layer
- 1. A source of record for public claims. Pricing, scope, certifications, policies and service areas live on plain, current, structured pages with a named owner and a review date. Ask: if an assistant quoted your pricing page today, would the answer be correct?
- 2. A crawler access policy. Decide which AI crawlers may read the site, what they may take, and how often, then monitor against that decision. Ask: do you know which AI crawlers visited last month and what they read?
- 3. A recurring answer-engine audit. Query the major assistants on your company, products and category on a fixed schedule and record what they say. Correct errors at the source page they draw from. Ask: when did anyone last check how the leading assistants describe your company?
- 4. Claims traced to evidence. Every security, compliance or performance claim links to proof a buyer can request or inspect, such as a SOC 2 report, a certification record or published case data. Ask: can a prospect trace each claim on your security page to a document?
- 5. A fast path to confirmation. A buyer arriving from an AI answer came to confirm a decision. Give them the proof in one screen, without interstitials or pop-ups in the way. Ask: how many clicks does it take to confirm your core claim from the homepage?
How we approach it at Aegis
We treat a client's public content as a governed data source, held to the same evidence discipline as their SOC 2 and AI governance programs. That starts with an inventory of the claims the site makes and who owns each one, then a review of how the major assistants currently describe the business. From there we set the crawler policy and the review cadence, and sequence the fixes so marketing keeps publishing while the record gets accurate.
Let the answer engine start the conversation. Make sure your website can finish it.