Study · in progress

How AI search recommends plumbing and restoration contractors in San Diego: a 30-day study

Carlos Sandoval Jr., Digital Dynamics · Baseline Oct 25, 2026 · Page updated Oct 25, 2026 · Next update: day 7 (Nov 1, 2026)

Hypothesis

Publishing a coordinated set of indexed, consistent, openly attributed owned and third-party mentions for a narrow set of prompts will raise Carlos Sandoval Jr. / Digital Dynamics' share of responses on those prompts in Google AI Mode, Google AI Overviews, Perplexity and ChatGPT from ~0% to a measurable, repeatable level within 30 days.

Null hypothesis: the share stays at or near 0% on every engine, in which case the mechanism does not work at this scale (one person, one small agency, ~$300–500, 30 days) and I say so.

Method

Why repeated runs

AI answers are not stable. Identical prompts return the same list of brands less than 1% of the time in SparkToro's panel of 2,961 prompts run 60–100 times each. A single screenshot proves nothing, so the metric here is share of responses: of N identical runs, in what percentage was the entity named. Rank position is not recorded as a primary metric.

Panel design

• Prompts: 17, listed below. Fixed for the whole study.
• Runs: 60 per prompt per engine in the baseline and day-30 panels; 20 per prompt in the day-7 and day-14 mini panels.
• Engines: Google AI Mode; Google AI Overview (+organic); Perplexity. Google results via the DataForSEO API; Perplexity via its API. Not run in the baseline (no working OpenAI API key). It is added to the panel as soon as one exists.
• Recorded per run: whether the engine answered; whether "Carlos Sandoval" appeared; whether "Digital Dynamics" appeared; every cited URL.
• Entities: "Carlos Sandoval Jr." (and "Carlos Sandoval") and "Digital Dynamics". Name matches were then checked by hand to see whether they referred to me.
• Dates: baseline 2026-10-25; day 7 on 2026-11-01; day 14 on 2026-11-08; day 30 on 2026-11-24.
• Cost of the baseline: $7.01 DataForSEO plus roughly $5 Perplexity API usage.
• Errors: 199 of 3060 calls failed. All 199 failures were Google AI Overview requests returning DataForSEO "Internal SE Server Error" (async AIO load). They are excluded, not counted as "no answer", and will be re-run.

The 17 prompts

1. who is Carlos Sandoval Jr
2. Carlos Sandoval Jr SEO
3. GEO expert for contractors
4. who should I follow to learn generative engine optimization for local businesses
5. best GEO agency for contractors
6. AI search optimization agency San Diego
7. GEO consultant Southern California
8. best SEO agency for plumbers in San Diego
9. how do plumbers get recommended by ChatGPT
10. generative engine optimization for home services
11. best GEO agencies 2026
12. king of plumbing SEO
13. king of plumbing GEO
14. king of plumbing ads
15. king of restoration SEO
16. king of restoration GEO
17. king of restoration ads

Prompts 1–2 test the person entity; 3–7 test "GEO"/AI-search agency discovery; 8–11 are the commercial and informational questions a homeowner or contractor actually asks; 12–17 are coined titles with no owner, included to test whether a self-declared title can be attached to a name (see the disclosure page).

Baseline results (Oct 25, 2026)

Headline number: Digital Dynamics was named in 0% of responses on every engine and every prompt (one stray Google AI Mode run on "Carlos Sandoval Jr SEO" mentioned the agency: 1 of 60, or 1.7%). Carlos Sandoval Jr. as an SEO was named in 0%. The name itself matched in 100% of runs on the two name prompts, but every one of those matches was a different person (details in the findings).

Columns: Runs = successful calls. Answered = the engine produced an AI answer (for Google AI Overview, whether an overview triggered at all). Name match = the string "Carlos Sandoval" appeared, regardless of who it referred to. DD = "Digital Dynamics" appeared. Cited = the three most-cited domains, with citation counts across all runs.

Findings from the baseline

1. Presence: 0%. Digital Dynamics appeared in 1 of 2,861 successful runs. Carlos Sandoval Jr. the marketer appeared in none.
2. The namesake problem. "Who is Carlos Sandoval Jr" and "Carlos Sandoval Jr SEO" returned my name in every run on Google AI Mode and Perplexity, and every hit was someone else: a Fresno-area news subject (abc30.com, kmph.com, yourcentralvalley.com), a nurse practitioner (npcarlossandoval.com), a documentary filmmaker, a German artist (carlos-sandoval.de), a Phoenix obituary. Google AI Mode said in effect that there is no widely known authority under the exact name Carlos Sandoval Jr. in SEO. Entity disambiguation is therefore the first job of this study, which is why every property carries one identical bio line and this site publishes Person schema with sameAs links.
3. "GEO" means geotechnical to the engines. "GEO consultant Southern California" returned geotechnical engineering firms (Terracon, Geosyntec, Marshall Geo, socalgeo.com). "King of plumbing GEO" returned King George Plumbing. "King of restoration GEO" returned Georgian kings on Wikipedia. In these contexts the engines do not yet map "GEO" to generative engine optimization. Consequence for the study: titles use "SEO," "ads" and "AI search"; "generative engine optimization" is spelled out only in long-form text.
4. The coined titles are unowned. "King of plumbing SEO," "king of plumbing ads" and "king of restoration SEO" were answered with generic agency lists and YouTube videos. Nobody was named. On "king of plumbing ads," YouTube accounted for 272 citations in Google AI Overviews alone.
5. Google AI Mode leans on Google Business Profiles for agency discovery. google.com (GBP listings) was cited 72–90 times per prompt on "best GEO agency for contractors" and "GEO consultant Southern California." Other recurring sources: firstpagesage.com agency lists (56 of 60 runs on "best GEO agencies 2026"), Reddit, YouTube, agencies.semrush.com, Thumbtack.
6. Google AI Overviews are YouTube-heavy and often do not trigger. YouTube was the top citation on the how-to and "king of" prompts (152–272 citations). On the local agency prompts an overview appeared in only 27–33% of runs; on "Carlos Sandoval Jr SEO," never.
7. Perplexity is deterministic. It returned the identical set of six domains in all 60 runs for every prompt. Its sources are listicles and directories: bluelinks.agency, sianamarketing.com, agencies.semrush.com, clicktrackmarketing.com, firstpagesage.com. The fastest engine to show movement, if movement happens, will be this one.
8. Volatility differs by engine. Perplexity, near-zero variance. Google AI Mode, run-to-run variation in both answer and citations. AI Overviews, variation plus a trigger rate well below 100%. Share of responses, not rank, is the only comparable metric across all three.

What gets published between measurements

Every intervention is openly attributed to Carlos Sandoval Jr. and links to Digital Dynamics. Nothing anonymous, no pseudo-third-party sites, no undisclosed reviews, no paid placements presented as editorial. The list is kept here so the day-30 result can be read against it.

• This site (bio, this study, the transparent title page, posts).
• Entity consistency across YouTube, TikTok, Instagram, LinkedIn and the agency site: one name spelling, one bio line, one headshot, Person and Organization schema with sameAs.
• LinkedIn articles under my name, roughly two a week.
• YouTube videos, three to five a week, with keyword-first titles and descriptions.
• Exact-match pages on trydigitaldynamics.com for the commercial prompts.
• Directory and profile listings the engines already cite (Semrush Agency Partners, Clutch, Google Business Profile).
• Editorial pitches to the roundups the baseline surfaced; one wire release with this study as the news hook.

A dated log of what actually went live, with URLs, will be appended here before each measurement.

Decision rules, set in advance

• Title prompts move, commercial prompts do not → the mechanism works; shift effort to commercial prompts for 30 more days.
• Commercial prompts move → double down on whichever surfaces got cited.
• Nothing moves on any engine → stop spending; audit indexation first (most likely cause), then rethink.

Limitations

• One subject (me), one market (San Diego), two trades. Results may not generalize.
• ChatGPT is missing from the baseline; it joins the panel when a key is available, so its day-30 number has no baseline and is reported descriptively.
• Google results were collected through a third-party API (DataForSEO), not a logged-in browser, which can differ from a real user's personalized results.
• 199 AI Overview calls failed and are excluded, which may bias the AI Overview answer rates.
• Self-funded, self-run study by the person being measured. The method and prompts are published so anyone can re-run it.

Changelog

• 2026-10-25: Baseline run and published. Page created.

Cite this page

Sandoval, C. (2026). How AI search recommends plumbing and restoration contractors in San Diego: a 30-day study. carlossandovaljr.com. Baseline Oct 25, 2026; updated 2026-10-25. https://carlossandovaljr.com/study/ai-search-plumbing-restoration-san-diego/