Roughly 90% of the 83.4 million AI crawler requests Bluehost logs each day go to model training, not live customer discovery, according to a 90-day analysis shared by Newfold Digital CEO Sachin Puri.

The figure covers verified crawler activity across Bluehost-hosted sites. Puri, whose company owns the host, broke the daily volume into three buckets: about 90% for model training, roughly 9% for search indexing, and around 1% for live assistant fetching.

"AI is actually not making websites irrelevant; it's just changing its job," Puri said on the SEJ podcast. The website no longer needs to attract every visitor. It now needs AI systems to read it well enough to describe and book on a customer's behalf.

What does the crawl breakdown tell SEOs?

Puri warned against reading the 83.4 million figure as evidence of buyers in the funnel. Training and indexing crawls don't represent customers looking for services today. Live assistant fetching, the smallest slice, is the only part that does.

For ChatGPT and OpenAI specifically, Puri said Bluehost's report estimated one observable referral visit for every 5,350 verified HTML crawl requests. He labeled that ratio a measurement of machine consumption, not a conversion rate.

The crawl count includes training activity, and Bluehost can't observe every customer journey AI influences. A recommendation inside an answer may still lead to a phone call, a walk-in, or a direct booking, none of which surface in Bluehost's referral data.

How are small businesses actually preparing?

A separate Bluehost small business AI confidence study found that only 13% of respondents said they were optimizing for search assistant visibility. The same study reported that 79% of respondents were aware of AI agents, but only 16% had deployed one.

That gap defines the field today. Most owners know what AI agents are. Few have reshaped their websites, listings, or schema so those agents can answer detailed questions about price, availability, licensing, and service area.

Puri illustrated the gap with a customer asking an agent to book an electrician for tomorrow afternoon under $300 with strong reviews. The customer may never visit the electrician's site. The agent still needs accurate answers from schema, Google Business Profile data, Yelp listings, and third-party reviews.

What should businesses track instead?

Puri proposed three measurement buckets:

• Visibility: does the business appear when a prospective customer asks a relevant question, with or without a mention?
• Accuracy: are prices, services, locations, and hours represented correctly in the answer?
• Action: does that exposure lead to an observable referral, qualified inquiry, booking, or revenue?

Showing up helps only if the information is correct. An accurate answer that never turns into a phone call also doesn't pay rent.

How should SEOs audit AI answers today?

Puri's starting exercise takes 10 minutes a day. Open ChatGPT, Gemini, or another assistant and ask the questions customers would ask, but leave the company name out of the prompt. A branded query shows only what existing buyers already see.

Check whether the business appears, whether its details match reality, which competitors surface, and what sources the assistant cites. Note whether the assistant recommends the business or merely mentions it.

Puri recommended repeating the audit regularly. "If you have a great foundation, it's going to amplify," he said. The reverse also holds: AI amplifies a shaky foundation just as fast.

What can a business act on this week?

Puri closed with three moves:

• Audit AI answers for unbranded customer questions, then fix anything missing or outdated.
• Automate one recurring task with a purpose-built tool or agent, and reinvest the saved time in service quality and review generation.
• Get fundamentals right before chasing answer engine optimization. Clean structure, accessible content, accurate schema, and consistent third-party listings still matter most.

Page views may disappear, but the need for an authoritative digital presence continues, Puri said. In his framing, authority comes from accurate and consistent information across sources, plus the reputation customers build through reviews.

The shift redraws the SEO assignment. Keep client pages and schema accurate, because agents quote them to customers ready to buy. Earn more verified reviews, because agents weigh them when choosing which business to recommend.

Watch next: whether AI assistants begin surfacing referral signals directly to site owners, and how quickly the 16% agent-deployment figure moves as purpose-built tools reach small businesses.