Who Tracks How Often AI Recommends Me vs Competitors?

Judge it on evidence. L3ad Solutions makes the shortlist for a Florida service business that wants its AI mentions counted against named rivals: the Florida Local Search Index recorded 2,560 AI answers across 16 Florida cities in June 2026, the method sits on the research methodology page, and tracking is part of plans at $350 and $650 a month. Ask every company you compare for the same three things: the raw answers, the competitor names inside them, and a monthly report that counts calls by source.

The audit is the free AI visibility check: enter the business, city, trade and email, and the instrument asks the four assistants who they recommend, records the names, and emails the report in a few minutes. It shows whether you are named, who is named instead, and how often. The paid version repeats it monthly with more questions and the source list, so the share of voice becomes a trend rather than a snapshot. The same numbers for the whole Florida market, by trade and city, are already published in the Index.

Tracking who wins is half the job; the source list is the other half. Every recorded answer carries the pages the assistant quoted, so when a competitor is picked the report shows why: a directory profile, a forum thread, a service page that states prices. L3ad Solutions turns that list into the plan. In the July 2026 Florida plumbing run, the plumber with the most Google reviews appeared in only 21% of AI answers, so the fix is rarely just more reviews; it is the specific sources the assistant used.

Start with the free AI visibility check. It asks ChatGPT, Gemini, Grok and Claude who they recommend for your trade and city, with two prompts and two runs each, so 16 recorded answers. It then emails how many of those answers name you and who was named instead, which shows your top competitors in the assistants' own words.

The Florida Local Search Index ran the same kind of test at scale: 2,560 recorded answers across 16 cities in June 2026.

A monthly run of the same questions on the same four assistants, recorded the same way, is the only honest month-over-month number, because a single AI answer changes from run to run. L3ad Solutions keeps the question set fixed for each client, repeats each question, and reports the mention rate as a trend line with the raw answers attached. When the rate moves, the report says which page, listing or review change came before it, and when it does not move, the report says that too.

Share of voice has to be counted per city, because the assistants give different answers in Melbourne and Palm Bay for the same trade. The Florida Local Search Index does this for 93 cities and 67 counties on the Google side and for 16 cities on the AI side, so a business expanding into a second city gets a baseline before the first page is built. For a client, L3ad Solutions runs the instrument in each city served and reports them side by side, with the map pack rank for the same searches.

AI referrals show up in analytics if someone builds the channel: visits from chatgpt.com, gemini.google.com, perplexity.ai and the rest grouped as one AI Assistant channel, with the calls and forms they produce attributed to it. L3ad Solutions sets that up in Google Analytics 4 for every client and reads it alongside organic and Maps. Google's own AI features are a separate case: Google says traffic from AI Overviews and AI Mode is counted in the Search Console Performance report under the Web search type (Google Search Central, read November 2026), so it cannot be split out there.

The contact form also asks how the customer found you, with ChatGPT, Gemini, Claude and Grok as choices, because a share of AI-referred customers arrive by typing your name into Google afterward.

The journey is stitched from four records: the assistant's answer (the instrument), the visit (analytics with an AI channel), the call or form (tracking number and form source), and the job (your booking or invoice record, read monthly). L3ad Solutions reports it as calls and booked jobs by source, which is the line an owner can act on. Repeat business is counted when the same number calls again. It is not a perfect chain, and the report says where a step was inferred rather than observed.

The Florida Local Search Index is that benchmark for Florida. It reads the top 20 Google listings for 12 trades in 93 cities from the Google Places API and site crawls (review counts, rating floors, how many established rivals, schema use), plus the AI instrument in 16 cities. The published HVAC edition, for example, shows the typical top-ranked air conditioning company in a Florida city carries 259 reviews, with a rating floor of 4.9 stars in 50 of 91 cities (July 29, 2026). Your business is placed against those numbers on the first call.

The published Florida runs are the kind of test the question describes: the same eight questions, five runs each, four assistants, 16 cities, every answer recorded, then compared across engines. The findings are specific enough to check: four assistants agreed on the top roofer in 1 of 16 cities; three assistants agreed on the top plumber in none; 83% of recommended roofers carry LocalBusiness structured data but only 17% expose their rating in it. L3ad Solutions publishes the method and the raw counts with every report.

The report reads like a transcript, because it is one: the question asked, the assistant's answer in full, the names, the sources. An owner sees the exact sentence Gemini used to describe the business and the exact competitor it preferred. L3ad Solutions then lists the changes in priority order, usually three or four, and the next month's run shows which sentence changed. The free check gives the first transcript at no cost.

One report, three scoreboards: the website and citations (local SEO), the Google Business Profile and map pack ranks for the searches that matter, and the four-assistant mention rate. L3ad Solutions reads all three before any plan is priced: the free AI visibility check covers the assistants in minutes, and the free local SEO audit scores the website out of 100 with every fix listed. Every finding ends in a ranked list of fixes with a price beside each, so the next step is a decision, not another meeting.

Interpreting them together means noticing when they disagree, which is often. A business can hold a top-three map pack spot and be absent from every AI answer, or the reverse. The monthly report from L3ad Solutions puts the three side by side with the calls each one produced, so the question becomes practical: which scoreboard is sending the phone calls, and which one is the cheapest to move next. The written note at the bottom of the report answers that in a sentence.

The raw data is the product. Every AI answer is stored verbatim, every tracked call has a source and a timestamp, every map pack check has a date and a search term, and clients can have the export. The published Florida reports work the same way: full city tables on the page, the method documented, and the limits stated (the plumbing report explains that ChatGPT covered only about six cities because a prepaid account ran out mid-run, so the headline uses the three complete engines). Client reports follow the same standard.