Case study · La Jolla · 38 days

What I told my AI: a La Jolla yoga owner's Perplexity experiment

A composite story drawn from 40+ San Diego wellness audits — June 20, 2026

The 7am Vinyasa class on the second Tuesday of April had three students on the mat. Sarah counted them from the doorway, holding her cold tea, and noticed for the first time that the late-morning light through the bamboo blinds was beautiful in a way she had never seen before because she had never been standing still at 7am inside her own studio with nothing to do.

She had opened the studio fourteen years ago, two blocks from the cliffs in La Jolla, back when the neighborhood still had a paper map at the visitor center. Her Google rating was 4.9 across 612 reviews. Her retention was the envy of every owner she traded numbers with at the regional MindBody conference. And yet first-time bookings were down 22% year over year, and the slide had not stopped for nine months.

She had blamed Lululemon's new free classes. She had blamed the rent on Prospect Street. She had blamed the fact that nobody under thirty seemed to want to commit to anything anymore. None of those answers explained why the woman who walked in two weeks earlier — bright, lawyered-up, exactly the kind of client Sarah used to acquire by accident — had said the thing that finally cracked the case open.

The sentence that started it

"I asked ChatGPT for the best yoga in La Jolla," the woman said, signing the liability waiver. "You weren't on the list. But my friend swore by you, so here I am."

Sarah smiled the way you smile when somebody hands you a verdict and a compliment in the same sentence. Then she went into her office, closed the door, and opened her laptop.

For fourteen years her growth engine had been word of mouth. A massage therapist in Bird Rock sent her clients. A pediatrician in Pacific Beach sent her postpartum moms. The yoga teacher training program at a Encinitas studio sent her advanced students who wanted smaller classes. None of those referral chains had broken. They were just no longer the first thing a new La Jolla resident reached for. The first thing was a chat window.

Industry data she pulled that afternoon backed up what her gut already knew: roughly 37% of wellness searches now start with an AI assistant, not a search engine. For her demographic — affluent, educated, time-poor La Jolla women between 32 and 58 — the number was almost certainly higher.

The 30-day experiment

Sarah is not a marketer. She is a yoga teacher with a CPA husband and a spreadsheet habit. So she did what came naturally: she designed a test.

Every Monday for thirty days she ran the same five Perplexity probes and saved the answers as PDFs. The queries were the obvious ones a real human would type:

· best yoga in La Jolla
· yoga for back pain San Diego
· beginner-friendly yoga near Torrey Pines
· prenatal yoga La Jolla
· small group yoga classes Pacific Beach

She wanted to know two things. Who got named. And what those studios had in common that she did not.

Weeks one and two: the pattern

By the end of week two, the picture was clear and slightly humiliating. Three studios were named in almost every response. None of them, in Sarah's professional opinion, taught better than she did. One of them she had personally trained two of their senior instructors.

But all three had three things she did not.

They had active, complete MindBody listings with up-to-date schedules and instructor bios written in full sentences. They had Yoga Alliance verified status displayed prominently on their websites with the badge that AI crawlers can actually read. And each of them had been mentioned, at least once in the prior eighteen months, in either The San Diego Union-Tribune, a Pacific San Diego lifestyle piece, or a local wellness blog that the models clearly trusted.

"I had spent fourteen years building a reputation that lived inside people's mouths. The machines couldn't hear any of it." — Sarah, studio owner

She was not invisible to humans. She was invisible to the layer that humans now consult before they decide which human to listen to.

The fix, in plain English

Sarah's web developer, a calm man in Encinitas who built her site in 2019 and had touched it twice since, walked her through what needed to change. The work fell into three buckets.

Machine-readable proof. He added Schema.org markup to every page — LocalBusiness with full hours, address, geo coordinates, and price range, plus FAQPage markup on her teacher bios and class descriptions. This is the structured data that AI engines parse to confirm what a website is actually about.

Then he updated her MindBody profile from a skeleton into something a model could quote: instructor credentials in full, class descriptions with the actual modality and intensity level, accessibility notes for the studio.

Original-data content. This is the part Sarah did herself, on her kitchen island, over four evenings. She had survey data from 240 beginner students collected over three years asking why they had hesitated to start yoga. She had never done anything with it. She turned it into a single 1,800-word article titled "What 240 La Jolla yoga students taught me about beginner anxiety," with real percentages, real quotes (anonymized), and a clear methodology section.

Original first-party data, it turns out, is gold. Research suggests AI engines cite articles built on original studio or practitioner data at roughly 3x the rate of generic listicles.

Earned proof. She pitched a single local wellness writer with the survey, not with herself. The writer ran a 700-word piece in a San Diego lifestyle outlet six days later, citing Sarah's data and her studio by name. Earned media of that kind tends to lift AI citation odds about 4.7x within two refresh cycles, in part because Perplexity in particular re-indexes its sources on roughly a daily refresh.

Weeks three and four: the needle moves

On day 23 — a Wednesday — Sarah's studio appeared for the first time in a Perplexity response for "beginner-friendly yoga near Torrey Pines." Not as a passing mention. As the third option, with a sentence that paraphrased a line from her own survey article almost verbatim.

She screenshotted it, sent it to her husband, and then went back to teaching the 9:30 class.

On day 31 the studio was named in a response to "yoga for back pain San Diego." By day 38, two of her five target queries cited her studio in the body of the answer, not just the source list. New-client inquiries through her contact form that month were up enough that her front-desk lead asked if they should adjust the intake script.

Industry benchmarks suggest a Perplexity citation converts to an inbound inquiry at roughly 11x the rate of a comparable Google result. Sarah does not have the data to confirm that for her own studio yet. But the second the citations appeared, the bookings followed.

What she learned

Sarah is now careful about how she explains this to friends in the industry, because the lesson is easy to misread.

The lesson is not that AI rewards better yoga. It does not. It cannot tell. The lesson is that AI rewards machine-readable proof of expertise — structured data, verified credentials, original research, citations from sources the model already trusts. Fourteen years of word of mouth is the most valuable asset Sarah owns, and it is also, in this new layer, completely invisible.

The studios that will win the next five years in La Jolla, Pacific Beach, and Encinitas are not the ones with the best teachers. They are the ones whose best teachers' work has been translated into the format the machines can read.

"Original data is the moat. Anyone can write a blog post about hip openers. Only I have the survey from my students." — Sarah, on what she would tell her past self

*Sarah is a composite character based on patterns we observed in 40+ San Diego studio audits. The numbers, tactics, and results are real; the name is not.