AI Assistants Sent 5.6% of Local Business Website Form Leads in 2026: Savo Group Lead Study

What Savo Group Found in 2026 Lead Data

AI assistants such as ChatGPT sent 5.6% of the website form leads that 18 local businesses received between June 1 and November 12, 2026. That is 55 of 983 deduplicated form leads, measured in the CRM where each lead was recorded. Each finding below is written to stand on its own, with its sample and window.

• AI assistants sent 5.6% of website form leads (55 of 983) at 18 local businesses from June 1 to November 12, 2026.
• For home-service contractors, the AI share was 7.3% of form leads (47 of 644 at 14 businesses) in the same window.
• Every one of the 11 businesses with 10 or more form leads received at least one AI-referred lead, and the median business got 7.5% of its form leads from AI.
• ChatGPT sent 85.5% of AI-referred leads (47 of 55); Gemini sent 3, Copilot 2, Perplexity 1, Meta AI 1 and Claude none.
• 45.5% of AI-referred leads landed first on an inner page such as a service or city page, compared with 25.9% of Google organic leads (436 leads).
• Among the same 8 businesses, the AI share of form leads went from 3.6% in July 2026 to 7.0% in October 2026, a directional trend on small monthly counts.
• In GA4, AI assistant visitors completed a key event 8.2% of the time (25 of 304 sessions) against 3.1% for Google organic, across 4 businesses in 2026.
• AI crawlers made a median of 229 requests per site per day across 18 local business websites from November 4 to 12, 2026, and 15.2% of those requests were live fetches for a user’s question.
• Of 68 “best agency” list articles that AI assistants cited on November 13, 2026, 75% included the publisher’s own company and 40% ranked it first.

How We Measured AI Assistant Leads

Every lead in this study is a record in Persilio, the lead platform Savo Group runs for its clients. When someone submits a website form, Persilio stores the referrer, any UTM tags, and the first page of that session. We read the channel from those three fields at the lead level, not from analytics sessions.

The sample is 24 client businesses, 19 of which logged leads, for 2,339 deduplicated leads in total. The main window is June 1 to November 12, 2026, the period when form attribution was live on most sites. Inside that window, 18 businesses received 983 website form leads. Those form leads are the base for the headline figure.

We excluded anything that is not a real prospect. That covers deleted records, spam, job applications, waitlist sign-ups, QA and demo accounts, and Savo Group’s own leads. When the same contact came back in the same month, we counted one lead.

A lead counts as AI-referred when its referrer or UTM source is an AI assistant domain, such as chatgpt.com, gemini.google.com, copilot.com or perplexity.ai. Default channel rules had filed 2 Copilot visits as referral and 2 Gemini visits as organic. We counted all four as AI, by referrer.

Landing pages were sorted into types by URL: homepage, service page, city or location page, contact or quote page, pricing, blog, case study and about. Industry labels come from each business’s primary service. Only home services had enough businesses to report on its own.

AI Assistants Sent 5.6% of Local Business Website Form Leads

Organic search is still the largest source of website leads, and AI assistants now sit behind paid ads and ahead of referrals and social. For every AI-referred form lead, the 18 businesses received about 8 from organic search (447 against 55). That ratio is the clearest way to size the channel today.

AI leads were spread across the client base, not concentrated in one account. All 11 businesses with at least 10 form leads received one or more AI-referred leads. The median among those 11 was 7.5%. The single business with the most AI leads accounted for 25.5% of them (14 of 55).

One large account pulls the overall average down. The business with the most form leads in the study gets few from AI. Leave it out and the remaining 17 businesses got 7.2% of their form leads from AI assistants (52 of 723).

Window June 1 to November 12, 2026. Other industries in the sample had one business each, too few to report separately.

ChatGPT Sent 85.5% of AI-Referred Leads

When a local business gets a lead from an AI assistant, it almost always comes from ChatGPT. ChatGPT referred 47 of the 55 AI leads. Gemini, Copilot, Perplexity and Meta AI together referred 7. The first AI-referred lead in the data arrived in June 2026.

Crawling does not equal referring. Anthropic’s crawlers made 17.2% of AI bot requests to these sites in November (see the crawler section below), yet Claude sent no leads. Meta’s crawler was the busiest of all, and Meta AI sent one. For a local business, ChatGPT is the assistant that turns into form fills today.

AI Visitors Landed on Service and City Pages More Often Than Google Visitors

ChatGPT and the other assistants send people deeper into a website than Google organic search does. 45.5% of AI-referred leads began on a page other than the homepage. For Google organic, the figure was 25.9%. Assistants tend to link the page that answers the question, such as the service or the town, instead of the brand’s front door.

Service pages took 25.5% of AI leads against 12.2% of organic leads, and city pages took 12.7% against 8.5%. No AI-referred lead started on a blog post or case study. With only 55 AI leads, that zero is a pattern to watch rather than a rule, but it fits what we see in AI answers: assistants recommend businesses through their service and location pages.

Paid ads are shown for context. Their landing pages are chosen by the advertiser, so they are not comparable with the two earned channels.

The AI Share of Leads Rose Between July and October 2026

Among the same 8 businesses tracked for four full months, the AI share of form leads went from 3.6% in July to 7.0% in October 2026. We hold the group fixed so new clients joining the platform cannot create a trend on their own. The monthly counts are small, so read the direction, not the exact slope.

The wider sample points the same way. Split into two-month blocks across every business with form leads, the AI share was 4.6% in June and July (12 of 260), 5.6% in August and September (23 of 410), and 6.4% from October 1 to November 12 (20 of 313).

The business mix grew from 8 to 18 across those blocks, which is why the fixed group above is the better trend measure. November 1 to 12 came in at 3.8% (4 of 104). Twelve days of a small count is too short to call a reversal.

How Leads From ChatGPT and Other AI Assistants Behave

People who arrive from an AI assistant tend to write more and arrive during business hours. 90.9% of AI-referred form leads included a written message (50 of 55), compared with 82.6% of organic leads and 57.9% of paid leads. The median AI message ran 155 characters against 126 for organic.

30.9% of AI-referred leads came in after hours, meaning before 8am, after 5pm or on a weekend in the business’s time zone. For organic leads it was 43.8%, and for paid leads 54.8%. Our read is that people often use an assistant while researching at work, then fill in the form from the same session.

GA4 undercounts AI traffic, which is why this study counts at the lead. At the same four businesses from June to October 2026, AI assistants were 0.4% of GA4 sessions but 4.3% of CRM form leads. At one of them, the CRM recorded 12 AI-referred form leads while GA4 recorded zero AI sessions.

Consent banners, stripped referrers and channel grouping rules all hide AI visits from analytics. If you judge AI search from a GA4 traffic report alone, you will likely conclude it does nothing. Our guide to tracking leads from ChatGPT and Google AI Overviews shows how to capture the referrer on the lead record itself.

Which AI Crawlers Visit Local Business Websites

AI crawlers made 46,217 verified requests to 18 local business websites from November 4 to 12, 2026, a median of 229 requests per site per day. Cloudflare’s bot analytics identified each crawler by name. We grouped them by purpose using each company’s published documentation: training a model, building a search index, or fetching a page live because a user asked.

Meta’s training crawler was the largest single bot at 21.9% of requests. GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Amazonbot and DuckAssistBot reached all 18 sites. The live fetches matter most to a business owner: each one is an assistant reading your page to answer a real person, right now.

Google-CloudVertexBot made 16 requests on 2 sites. Googlebot is not listed because Google uses the same crawler for Search and AI features, so its AI share cannot be separated.

What AI Assistants Cite When Local Business Owners Ask for Marketing Help

When a business owner asks an AI assistant to recommend a marketing agency, the answer is mostly built from “best agency” list articles and directories. We ran this study to see which sources sit behind those answers, so owners can judge what they are reading. It is a snapshot of one day, not a ranking of agencies.

On November 13, 2026, we collected 112 answers with 694 cited links. We asked ChatGPT, Google AI Mode and the Perplexity Sonar API 25 agency-selection prompts and 8 marketing statistics prompts. Gemini answered 13 of the agency prompts. Prompts covered 8 general questions, 12 industry-specific questions, and 5 that named a Pacific Northwest city.

Questions that named a city were answered mostly from directories and Google business listings. Together they made up 59.5% of citations on city prompts (66 of 111). Semrush’s agency directory (13 citations) and Clutch (8) were the most-cited directories. For a local service business, the same pattern suggests your directory and Google Business Profile details feed straight into AI answers.

Statistics questions were answered from research and benchmark pages. 76 of the 131 citations on statistics prompts (58.0%) went to pages that published original data or benchmarks, and another 32 went to cost guides. Assistants look for a page that states a number, a sample and a date. That is the reason this study is written the way it is.

Three in four cited “best agency” lists were written by an agency that included itself. 51 of 68 listed the publisher, and 27 ranked the publisher first. Fewer than half (32) explained how they ranked anyone. That is a common publishing format, and it is worth knowing about when an AI answer leans on it.

The cited lists shared a format. 60 had the year in the title, 60 had a comparison table, 48 had an FAQ, and 47 had been updated within 60 days. The median list ran 3,643 words. Freshness and structure appear to matter to assistants as much as the publisher’s identity.

If you are choosing an agency with help from AI, ask for the evidence behind each name. Look for case studies with dated numbers, reviews on independent sites, and a clear method. The CSV linked below includes the full citation breakdown by question type.

What This Means for Your Local Business

AI assistants are a small but real lead source for local businesses in 2026, and the work that wins those leads is concrete. Here is what the data supports, in the order we would do it.

If you would like this handled for you, our AI SEO service covers page structure, entity details and lead attribution, and reports AI leads alongside Google leads.

Limitations of This AI Lead Study

This is an agency client base, not a random sample of US businesses. Every business works with Savo Group and most are in home services. Savo Group builds and optimizes many of these websites, so their AI share may run higher than a typical local business.

The counts are small. 55 AI-referred leads support the headline share and the assistant split. They do not support fine cuts by month, industry or page type, which is why we label the trend directional and report only one industry on its own.

Referrer-based attribution misses some AI influence. A person who reads a ChatGPT answer and then types the business name into Google is counted as organic or direct. Phone calls carry no referrer at all. Both effects push the true AI share up, not down.

The AI answer study ran each prompt once, on one day. AI answers change between runs and between users. Gemini answered 13 of 25 agency prompts and none of the statistics prompts. The crawler data covers nine days. We plan to rerun all three datasets and will update this page and its date when we do.

How to Cite This Research

You are welcome to quote, chart and republish these figures under the Creative Commons Attribution 4.0 license. Please credit Savo Group and link to this page. A short form that works in an article: “According to Savo Group’s 2026 lead study, AI assistants sent 5.6% of website form leads at 18 local businesses.”

Download the aggregate data as a CSV. It holds every figure on this page with its numerator, denominator, business count, window and source. It contains no client names, contact details or individual lead records. Questions about the method go to Taylor Rupe through our contact form.