Every agent eventually faces the same budget question: where should the next marketing dollar go? For decades the answer was geographic farming: pick a neighborhood, mail to it, and become a familiar name there. Today a growing number of consumers also ask AI assistants for recommendations, and agents are asking whether to spend money on being found there instead.
It is not an either-or decision. But there is a foundation to fix before spending on either.
The problem: sites that machines cannot read
NAR’s magazine reported in May 2026 that Rick Janson, a Compass agent in Denver, audited 50 top-performing agent websites and said he found more than 70,000 technical errors preventing them from being found by AI. “The issue wasn’t lack of content,” he said. “It was a lack of machine readability” (National Association of REALTORS®). Janson has also founded a company that sells AI-search services to agents, which is worth knowing when weighing the finding (National Association of REALTORS®).
What machine readability involves
- Clean text versions of pages, so an assistant does not have to guess which part of a page is the answer.
- Discovery files such as llms.txt that describe the site for automated readers.
- A precise statement of what you do and where you are licensed. For a Florida licensee this is a compliance point, not a detail: an AI summary that implies you can represent buyers outside Florida is a misleading advertisement built from your own site.
- Bot controls that admit legitimate crawlers while filtering forged ones.
Whoever does this work for you, the brokerage name and licence disclosures required by Rule 61J2-10.025, F.A.C., have to appear on the machine-readable versions of your pages just as they do on the visible ones.
What is reported about AI search
HousingWire, citing SparkToro’s analysis of Similarweb data, reported that 68% of U.S. Google searches ended without a click in early 2026 (HousingWire). The same NAR article describes Janson’s account of a client who found him through a Gemini search and became a $1.7 million listing appointment (National Association of REALTORS®). That is one agent’s anecdote, reported as such, and not evidence of what any other agent should expect.
HousingWire also reports that businesses cited in AI answers tend to have complete, consistent Google Business Profiles, and that inconsistent names, addresses and phone numbers across directories undermine that (HousingWire). For a licensee, that consistency includes the brokerage name.
What farming costs
One coaching firm puts a modest 500-home farm at roughly $500 to $2,000 a month depending on mail frequency and design (Power Unit Coaching). The arithmetic of a farm is simple to illustrate: in a hypothetical 1,000-home area where 4% of homes sell in a year, there are about 40 sales; your share of those is what the spend has to justify. That is an illustration, not a projection of what any farm will return.
The first version of this article repeated a vendor calculator’s projection of a 1,873% five-year return on farming. It has been removed: it is a sales tool’s forecast, and return forecasts are not something I will repeat.
Blending both
The two approaches reinforce each other. A neighborhood market update can be a mailer and a web page. A process explainer can answer a homeowner’s question at the door and an assistant’s question online (Neuhaus Realty Group). Keep neighborhood content about the place and the housing, never about who lives there.
How to start
- Audit your current spend between physical farming and digital visibility.
- Fix the foundation first: a site that machines can read, with your brokerage and licence stated clearly.
- Test small. Move a modest share of the budget for 90 days and track enquiries by source.
- Keep what works. If a farm produces listings, reinforce it rather than abandoning it.
- Check your listings everywhere. Same name, brokerage, address and phone across Google, Bing and directories.
- Write for questions. Make the question the heading and answer it directly (National Association of REALTORS®).
A final caveat
There is no credible public research yet isolating how much a machine-readability fix moves an individual agent’s lead volume. Anyone quoting a precise percentage is estimating. Treat the channel as worth testing, and measure your own results.