AI visibility in landscaping: measured by season, service and town

Named, recommended and first, for one service in one town

Every answer can get three verdicts, whatever the trade. Named: your company appears anywhere in the answer, under any form of its name a homeowner would recognize as yours. Recommended: the answer puts you forward as a choice for the job that was asked. First: yours is the first business the reader meets. Our method records two of them, named and first, and does not weight either: the visibility score is the share of the twelve questions in which the company was named, with first place set beside it.

A landscaping company needs one more thing beside each verdict: what it was for. An answer from an AI assistant names you for something: weekly mowing in one town, a paver patio in another, a sprinkler blowout, a plowing contract. You can be first for spring cleanups in your home town and missing from every irrigation answer two towns over, and one number for the month cannot say which. So the verdicts are written per line and read by service and town before anyone reads the total.

Write down the wrong answers as well as the missing ones. An answer that names you for aeration you do not offer, or for a town past the edge of your route, counts as named in the score and is still wrong, so the record marks it as both. Where a wrong town comes from is traced in the route-map check, by hand, in an afternoon.

Which lines only count in their own season

Your booking records say when each service's calls come in. A spring cleanup line read in August describes the answer to a question your office is not being asked that month, so it cannot tell you much about March. Of the twelve lines in the panel below, eight name a service that is bought in one part of the year.

So each line carries an in-season window, written beside it when the panel is set up: the months your records say that service's calls arrive. A spring cleanup line might count from [late winter] to [late spring], and a snow line from [late fall] to [the last snow] in a northern market and in no month at all where it does not snow. Take the months from last year's calls, and do not move them after a bad reading.

Out of season the line still runs, because the panel is fixed and a line dropped in spring and added back in fall has a hole where its history should be. What changes is how the answer is used. In season it counts toward the month's reading. Out of season it is a check for stale facts: an October answer that still gives your spring cleanup dates is repeating a page nobody changed after April, which is what the publishing calendar is for.

The lines with no season are the ones about the company itself: the question with your name in it, the comparison with a competitor, the small-job question. They count in every month, and they are the part of the reading the season does not move.

Twelve lines for one company, and the months each one counts

The template is for a company with one base and a route across several towns. Fill the brackets once, from your own services, towns and records, then leave the wording alone. Two towns are enough for the monthly panel, the one you work from and the route town you most want; the rest of the route belongs to the afternoon check against the map.

If HOA contract work is not work you want, swap that line for a second homeowner service before the first reading, not after it. Setting up prompt monitoring explains why a line reworded later starts a new series, and how each answer is asked and kept.

The service, the town, the price and the booking, beside every verdict

The published method keeps each answer with its verdicts, the other businesses named and the full text, pasted and dated. A landscaping company's record needs four more fields, because an answer can name the right company for the wrong job, put it in the wrong town, or quote a price the company never published.

The price field compares a number with a number: the answer's [price] per visit beside your page's [price]. When an answer quotes a price your pages never stated, the record is where you notice it, and the sources the answer shows, if it shows any, are where you look next.

The booking field is for companies whose routes fill. If your mowing routes close when they are full, a June answer telling a homeowner you are taking new lawns sends you a call you will turn down, and a March answer saying you are booked for spring when you are not loses you one. Both lead back to the line on your seasonal page that says where the season stands, and to whether that line carries a date.

Why May is read against last May

Two things move between April and May in a landscaping panel: whatever you changed on your site, and the season. In April the spring cleanup lines are in their window and the mowing lines are only arriving; by May it is the other way round, and the fall lines have not begun. If the pages the answers are built from get updated for the season, yours and everyone else's, part of any change from one month to the next belongs to the calendar.

The same month a year later holds the season still. May against last May reads the same lines in the same window, at the same point in the booking year, so what is left of the difference sits closer to what changed in between: your pages, your listings, a competitor who arrived. Month-to-month readings of the in-season lines still catch a sudden break, such as a page taken down, but the trend is the year-on-year line.

The first year has nothing to be compared with, so it is the baseline year, and every answer is kept verbatim for next year's comparison. Start before the season you care about most: a company that starts in February has a spring baseline by May, while one that starts in June waits nearly two years for its first spring comparison.

Keep a dated note of what changed on the site and the Business Profile, and of anything that moved the season itself: a late spring, a watering restriction that stopped mowing for three weeks, an early snow. A May that looks worse than last May can then have its reason written beside it, and a May that looks better can be checked against the changes you made.

Google's own report is dated as well. The generative AI performance report, rolled out in Search Console to all websites on August 31, 2026, counts impressions of your links in AI Overviews and AI Mode and splits them by page, country, date and device. There is no query dimension: it can show that the spring cleanup page appeared in April, not which question brought it up, so it sits beside the panel's spring lines rather than replacing them.

What one day's reading can show, and what it cannot

Our largest reading was a single day's. In the probe attached to our audit of 9,483 small business websites (40 trades, 90 metros in the US, Canada, Europe and the UAE, late August to early September 2026), one buyer question per business went to Claude with web retrieval on, and 14.9% of 9,240 businesses were named; 17.7% among the 5,928 answers that gave a differentiated list rather than a hedge. That is a Claude number from one question on one day, and we publish no landscaping cut of it.

It recorded names only: no service, no town, no price, no season. For a trade bought by the season, one day's answer reads whatever season that day fell in, so a company's first month on its own panel starts a record rather than judging a year. That first month shows where you stand: which services and towns name you, which lines never do, who holds the share of voice instead, and which prices and booking lines the answers already have wrong. What the site should state first is on the landscaping and lawn care page.

Having the panel run for you is monitoring, at $290 a month, which you can cancel in any month: twelve lines set up for your services, towns and months, put once a month to one assistant, Claude, and every answer kept. Doing it yourself is the same procedure in a spreadsheet, and it holds up as long as whoever reads the answers never rewords a line after a bad month.