How Futuro became the most-cited name in AI answers about salons.

Ask an engine which AI receptionist to buy and it answers from a deep bench: YouTube explainers, Reddit threads, companies with years of content behind them. Across thirty days of answers to the 28 broadest questions Futuro tracks, engines cited YouTube 917 times. They cited futurocorp.com 25 times, which put it ninetieth.

Narrow the same question to salons, and the order reverses.

Futuro didn’t get there by outspending anyone. It picked the questions where it had the better answer, made those answers readable by machines, and waited out the fourteen weeks it took.

Brandon Gillespie founded Futuro, which sells voice agents that answer a business’s phone and handle the calls a front desk would. In May he was trialling half a dozen GEO platforms at once and planning to pick one by the end of the month, with Trakkr, as he put it at the time, “between your company and two others”.

Trakkr asks a fixed list of questions across eight AI engines every day and reads the customer’s server logs alongside the answers. Futuro connected its Cloudflare logs on the first day, and they showed 123 requests from AI crawlers dying on a server error: a certificate fault between Cloudflare and the origin that no human visitor ever saw. Cloudflare was also set to its most restrictive mode, which challenges anything that looks automated. People got through. The crawlers that feed AI answers mostly didn’t, and in that first week no engine named Futuro in a single answer Trakkr recorded.

Both were cleared inside a fortnight. The certificate fault stopped within days, Cloudflare came off its strictest setting on 25 May, and in the three months since, AI crawlers have hit an error on two days.

An engine cannot recommend a page it cannot fetch.

On 2 June, Trakkr flagged something else in the same logs: every AI bot that requested /pricing was getting a 404. In a browser the page loaded normally, which is why nothing had flagged it. A person and a crawler don’t always see the same site.

The first fix didn’t take. The daily log check caught it the next morning, a second attempt went in on 12 June, and that one held. Twelve days later an AI answer linked to the pricing page for the first time, and Google’s AI Overviews, Perplexity and ChatGPT have all quoted it since. For its first month of tracking, the page built to answer the pricing question couldn’t.

Futuro calls what it sells human staff mirroring: a voice agent trained on one business until it answers the way that business’s own front desk would. Brandon coined the phrase, so in May it was a term with one company behind it and no presence in any AI answer.

Eight days into tracking, ChatGPT answered “human staff mirroring” with Futuro’s name. Perplexity followed two days later. From then on at least one engine named Futuro on every one of the 82 days that question ran, and two later variants of it were answered with Futuro too.

The bare term came out of the tracker in August and the comparison question stayed, because it is the harder one and Futuro was holding it.

A coined term is the easiest version of a larger idea. Ask an engine for the best AI receptionist and it chooses between Retell, Bland, Vapi, CloudTalk and a few thousand YouTube videos. Ask it what human staff mirroring is and there is one company to cite. Futuro’s work in June was making that source easy for a machine to read: FAQ and product schema went on in mid-June, and a longer guide that defines the term, compares it with receptionists and chatbots and says plainly what it can’t do earned its first citation on 3 July.

Most companies have a few of these questions without having invented a word for them.

Between 19 May and 19 August, Futuro completed 26 of the actions Trakkr raised. Each arrived as a dated recommendation with the page already drafted, rather than a score to interpret. Futuro’s writers, three of whom work on GEO full time, and its developers built and published every one through their own repository. Trakkr has never pushed a change to futurocorp.com.

AI crawlers came back more often as more of the site became readable.

By mid-August, GPTBot, ClaudeBot, PerplexityBot and the rest were reading futurocorp.com nearly three times as often as they had in mid-May, with the steepest climb after llms.txt went live on 23 July.

Not every win arrives on schedule. The nail-salon question was tracked from the first day, and for twelve weeks it produced nothing but the odd isolated day: on 2 June four engines named Futuro for it, and by the next day all four had moved on.

On 28 June Trakkr sent Brandon a mock-up, voice AI for nail salons, ranked, in the shape that was earning citations elsewhere. His team published their own version the next day. Two engines picked it up immediately, then let it go again for five weeks. Brandon checked for himself before any report caught up:

Perplexity cited the page on 6 August and has named Futuro in its answer every day since.

A single day means very little. A win counts when it holds, which is why Trakkr scores runs rather than days.

The wait got shorter after that. Three more beauty-salon questions went into the tracker in early August, and two of them named Futuro on their first day and have every day since.

The salon questions are held. The broad ones are the next target: best AI phone answering service, AI receptionist for inbound calls, and the rest still go to Retell, CloudTalk and YouTube. Futuro tracks them every day, and the plan for them is the one that has already worked twice. A page a crawler can read, a direct answer near the top, a comparison that names the competition, and then the wait.

AI assistants are still a small share of futurocorp.com’s traffic: 2.1% of sessions between 1 May and 22 August, four in five of them from ChatGPT. Those sessions have grown every month since tracking began, and on their own they wouldn’t be worth a page.

Then look at who asks for a quote. Of every session in those months that requested a quote or submitted the contact form, roughly one in eight came from that 2.1%. These are Futuro’s own Google Analytics figures, not Trakkr’s.

Someone who asks an assistant which receptionist to buy, reads the comparison and clicks through has already done the comparing. They arrive deciding when to call, not whether.