👋 Hey, Jon here!
This week, we are back to talking LLMs (large language models) and diving into Q1 data to examine trends in LLM-referred lead volumes, bookings, revenue, and more. This analysis also includes regional data to see if some areas produce higher volumes of LLM-referred activity.
Introducing The SearchLight Data Lab
For a long time, I’ve wanted to create interactive, visual benchmarks as a supplement to this written content.
I’m SO excited that this is now live and this article is now paired with a visual that will be updated monthly: ChatGPT and LLM Lead Trends for Home Services (2026 Benchmarks)
I’ve got also got another interactive view live: Regional Cost per Lead Benchmarks for Google Ads.
There are so many more to come, but I would so appreciate it if you got any value from this to share out the SearchLight Data Lab on social media, to your colleagues, teammates, and anyone else who might benefit from it.
With that, let’s dive into the data!
How We Track ChatGPT and LLM Performance
A brief note on methodology, so we’re all on the same page for this analysis. When ChatGPT or another LLM recommends a contractor and a consumer clicks through, the referral arrives with a direct fingerprint — utm_source=chatgpt.com in the URL, or a referrer: chatgpt.com header. Same for gemini.google.com, perplexity.ai, copilot.microsoft.com, and claude.ai.
There are cases where UTMs are ‘clipped,’ and it appears as if a referral is direct traffic to a site, so it’s important to remember that, even with completely dialed-in tracking, sometimes things fall through the cracks.
When available, we are looking at the direct traffic sources based on the UTMs (or call tracking if set up properly), which also provide us with other rich information about what content populates in LLMs:
For example, if an LLM populates a Google Business Profile or a Yelp profile, we’ll usually see that information in the UTM parameters.
In Q1, for example, 16% of leads had Google Business Profile markers in the referral path.
We also saw Yelp there, but at a much smaller volume.
This can create issues when trying to track these leads in your own systems, because GBP received credit in that scenario, even though the lead originated from an LLM. 👇
Only 1.5% of AI-Referred Leads Were Correctly Attributed in CRMs
Of the 2,337 AI-referred leads SearchLight identified in Q1, we checked what the contractor’s CRM recorded as the source for those same leads:
69.8% had no campaign attribution at all
28.7% were actively misattributed to a different source
1.5% were correctly attributed to AI
Only 1.5% of AI leads were correctly attributed across these CRMs.
And no, we are not picking on any particular CRM. We are integrated with 8 of them (and growing), and this sample includes accounts across all 8.
The majority of misattribution was a blank campaign, but within the 28.7% actively misattributed:
56.5% landed in catch-all campaigns such as ‘Main Website Number’ and others, suggesting a lack of call-tracking campaigns set up using LLM UTMs.
However, 51% of LLM leads converted via forms after reaching the website, and without proper form attribution, we saw a majority of those leads misattributed as well.
Attribution discrepancies are exactly why we built SearchLight, and frankly, I sometimes get shit for this data because I’m the bad guy calling out gaps. Marketing attribution is hard. Commercial accuracy lives in the details; there’s a lot of custom logic across ad channels, CRMs, conversion tools, websites, call tracking, etc. Please do not take this as a jab at CRMs; the reality is that getting clean data is hard, and it’s why we exist.
LLMs Generated 1.74% of Organic’s Lead Volume in Q1
In Q1 2026, AI chatbots, ChatGPT, Google Gemini, Microsoft Copilot, Perplexity, and Claude, sent 2,337 leads to contractors on the SearchLight platform.
Those leads produced 784 booked appointments, 356 paying customers, and $1.04 million in closed revenue.
That’s 1.74% of organic search volume. Small. We know.
But the quarterly number hides the more interesting story: how each month looked compared to the one before.
The data is telling me that direct lead volume from LLMs is low (relative to established channels). This does not surprise me, but what I’m watching very closely is how this activity is trending and the volume numbers for the top contractors in this sample. 👇
Booked Customers from LLMs Increased 44% from January to March
Q1 activity from LLMs wasn’t linear, as this chart indicates. We saw a dip in February, followed by a surge in March.
More leads, more booked customers, more paying customers, and more revenue opportunities generated (note that the sample size is consistent in all months).
Again, small absolute volume, but I am more focused on watching the relative shifts to look for signs of wider adoption and a possible acceleration.
The Adoption Curve Is Widening
As I personally try to measure the impact of LLMs on our industry, another metric I am very interested in is the % of contractors in this sample who received at least 1 lead from an LLM, which jumped over 7% from January to March:
In January, 29% of contractors on the SearchLight platform received at least one AI-referred lead. By March, that number was 36%.
Here’s the nuance: leads per contractor stayed flat at ~2 all quarter.
The growth isn’t coming from existing contractors getting more AI leads.
It’s coming from more contractors showing up in AI results for the first time.
This is cool to see, but the absolute lead volume of LLMs isn’t quite large enough to meaningfully move the average leads per contractor across a large sample, however 👇
Contractors Are Already Seeing Meaningful Volume from LLMs
The median contractor received 2 AI leads in all of Q1.
At the 90th percentile, that jumps to 7.
The top contractor on the platform received 37 AI-referred leads, representing 23% of their organic search volume.
While the distribution of LLM leads in this sample is heavily skewed, most contractors see very little, the top performers are seeing legitimate volume from LLMs:
LLM Lead Generation Regional Performance
It’s been asked over and over on Facebook groups and on other corners of the internet - ‘are some areas seeing more lead volume from LLMs?’ with hypotheses around the type of population (younger, tech savvy) and how it might influence the use of those tools in the search process.
Southwest and West contractors see about 35% more AI leads per account than the Midwest and Northeast. The per-contractor normalization suggests there’s a real adoption difference across regions.
At the state level, the top three by AI leads per contractor are Alabama (8.2), Georgia (7.8), and Texas (5.9).
The Southeast showing up this strongly was not what we expected. The assumption going in was that tech-forward markets out West would lead. The data says otherwise, at least for Q1.
Book Rate Is Improving
AI leads converted to booked appointments at a higher rate each month throughout Q1:
January: 31.5%
February: 33.6%
March: 35.1%
That’s a 3.6-point improvement over the quarter. Still below organic search (43.2%), but the gap is closing.
One possible explanation: as AI chatbots improve their local recommendations, the leads they send are arriving with stronger intent. A more specific recommendation (”call this company for a furnace replacement”) produces a higher-quality lead than a generic one (”here are some HVAC companies near you”).
Microsoft Copilot continues to have the highest book rate among AI sources at 41.1%, approaching organic-level quality. Small sample (95 leads), but it’s been consistent.
Revenue Opportunity Tells a Different Story Than Closed Revenue
Closed revenue didn’t improve like the other KPIs in the lead funnel: $398K in January, $273K in February, $364K in March.
When the volume is that low, a big job or two can make a noticeable difference, but closed revenue is the ultimate KPI, and it underperformed the rest of the funnel metrics in Q1.
Revenue opportunity generated from LLMs, on the other hand, looks different, with linear growth: $1.02M → $1.02M → $1.59M.
The March pipeline is 56% larger than January’s, but the conversion of revenue opportunities to closed revenue matters significantly.
Just 29% of the LLM-generated revenue pipeline converted to closed revenue in Q1, well below our industry average of 40% across paid and organic channels.
As I’ve written many times before, any metric measured in isolation doesn’t tell the full story. While top-of-funnel volume increases from these channels, LLMs underperform in converting those opportunities into closed revenue.
ChatGPT Still Generates The Majority of LLM Lead Volume
ChatGPT generated 83% of all AI leads and generated 10x the volume of Google Gemini. It’s the GBP of LLMs.
The below percentage growth of other LLMs looks impressive, but the volume is too small to make any meaningful insight out of this at the moment:
What This Means
Five things stand out from Q1:
1. The growth is in breadth, not depth. More contractors are being found by AI chatbots each month, but each contractor still only gets about 2 AI leads per quarter. What moved was the percentage of the network receiving AI leads at all: from 29% to 36%.
2. Lead quality is improving. The book rate climbed from 31.5% to 35.1% over the quarter. AI-referred leads are still lower quality than organic, but the trend is in the right direction.
3. Revenue pipeline is the metric to watch. Closed revenue is a lagging indicator on small samples. Revenue pipeline grew 56% and gives you the cleanest signal of where AI-referred revenue is heading, as long as a higher % of that pipeline converts to sold revenue!
4. Your GBP matters more than you think. ChatGPT is pulling from Google Business Profiles to generate local recommendations. If your GBP is thin, you’re probably not showing up in AI search results.
5. Your CRM is likely blind to this channel. 98.5% of AI-referred leads were either invisible or misattributed in the CRM.
We’ll update these numbers after Q2. Subscribe so you don’t miss it.
Explore the full interactive benchmark at searchlightdigital.io/chatgpt-llm-lead-trends-home-services.
Until next time,
Jon
Data: SearchLight Digital, Q1 2026. 707 home services contractors. AI channels tracked: ChatGPT, Google Gemini, Microsoft Copilot, Perplexity, and Claude. CRM misattribution analysis based on 2,337 AI-referred leads cross-referenced against CRM campaign fields. Geographic analysis based on 389 contractors with geo-coded leads.