News & analysis
AI Search Is Changing What Your Website Has To Do
Newfold Digital CEO Sachin Puri, whose company owns Bluehost, says AI assistants can now find a business and book it without the customer ever visiting the business's homepage, which shifts a website's job from attracting clicks to feeding AI systems accurate information.
By Roomer Editorial · · 4 min read

What's actually changing for websites?
An AI assistant can now search for a business that fits a request, confirm details, and complete a booking on its own. The customer may never land on the business's homepage.
Sachin Puri, CEO of Newfold Digital (Bluehost's parent company), discussed this on the Search Engine Journal podcast. His point: the website isn't becoming irrelevant, its job is changing. It's no longer just a page for humans to browse. It's the one part of a business's digital presence the business fully controls, so it becomes the reference point AI systems check against everything else they find — Google Business Profile, Yelp, reviews, forum threads.
Puri used an electrician as his example. A customer asks an agent to find someone available tomorrow, charging under $300, with strong reviews, and to book the slot. The agent has to answer the same questions a person would, by reading the website's content and structured data, then cross-checking it against other sources.
How much AI crawler traffic is happening, and what does it mean?
Bluehost ran a 90-day analysis across its hosted sites and counted 83.4 million verified AI crawler requests per day. Puri cautioned against reading that number as customer discovery.
Bluehost's internal breakdown attributed about nine in ten of those requests to model training, roughly one in ten to search indexing, and a sliver — about 1% — to live assistant fetching, the category linked to a real customer query in progress.
Specifically for OpenAI and ChatGPT, Bluehost's data showed roughly one referral visit turning up for every 5,350 HTML crawl requests it could verify. Puri stressed that number shouldn't be read as a conversion rate. Crawling, retrieval, citation, and an eventual booking are separate steps, and most of what gets crawled never becomes a visit a business can see in its analytics.
What should a business actually do about it?
Puri's framework has three parts: visibility, accuracy, action. Showing up in an AI answer doesn't help if the answer has the wrong price. An accurate answer still needs to convert into a call or a booking.
He pointed to a common failure mode: pages written to rank, loaded with vague phrases, light on facts. A page describing "reliable electrical services" gives an AI agent nothing to match against a specific request. A page listing services, coverage area, hours, licensing, and recent reviews gives the agent something to work with.
He gave a restaurant version of the same problem. A customer might ask an agent for a family-friendly restaurant with vegetarian and vegan options, wheelchair access, and a table for six that evening. The agent can't answer from a general description — it needs each attribute confirmed, and the booking made before another customer takes the slot.
Puri's suggested habit: spend ten minutes a day asking ChatGPT or Gemini the questions a customer would ask, without naming the business, and check whether it shows up, whether the information is right, and which competitors get recommended instead.
Bluehost's small business research found only 13% of respondents said they were optimizing for search assistant visibility, and 79% knew what AI agents were but only 16% had deployed one.
Roomer Editorial's take
Swap "electrician" for "boutique hotel in Lisbon" and the problem is identical. A traveler asks an agent for a pet-friendly hotel near Alfama with a rooftop and availability for three nights in June. The agent doesn't read your Instagram caption. It reads whatever structured, bookable information it can find and cross-check.
This is the argument for owning a page with real, current details instead of relying on a platform bio and a link in your profile. A Roomer site built from your videos gives each place you've covered its own bookable page — dates, pricing through rails like Vio, Stay22, GetYourGuide, Viator, or Klook, and the kind of specifics an AI agent needs to match a traveler's exact ask. A caption that says "amazing stay, 10/10" is the travel equivalent of "reliable electrical services." It tells a human nothing concrete and tells an AI system even less.
The crawler numbers in this story aren't about Roomer and shouldn't be read as a promise of referral traffic. The lesson is about structure, not volume: a page with real facts outperforms one built only to look good scrolling past it. Creators who treat their bookable pages as the accurate record of what a property actually offers are the ones an agent can act on. Creators still writing "stunning views, must stay" are optimizing for an algorithm that's already changing what it's grading on.
One move worth making this week: pick your five most-booked recommendations and check that the details on your own pages — amenities, pet policy, location specifics — are current and specific enough that an AI agent reading them could book with confidence.
Source: Search Engine Journal



