The visibility discipline
You are paying for "AI SEO" and you still cannot say whether ChatGPT or Google's AI Overviews actually mention your business, only what the agency's dashboard tells you to believe. Most firms selling AI SEO just added the letters AI to the same deliverables they shipped in 2019. Here is what actually changes when an agency treats AI answer engines as a second search surface, not a marketing slide, and how to tell the difference before you sign another contract.
What is an AI SEO agency?
An AI SEO agency works on whether an answer engine names your business, which is a different job from moving a keyword position. The check that separates a real one from a rebranded SEO shop is simple: ask for the prompt, the date it was run, and the answer that came back, then run it yourself. Anything that cannot be re-run by the client is an assertion, not a measurement.
Source: Semrush keyword research, verified in live pulls. Demand shifts; we re-verify before every engagement.
If your agency cannot show you the exact prompt and the exact answer, they are describing a feeling.
A working definition, not a sales pitch.
An AI SEO agency is a firm that manages both classic search rankings and how a brand appears inside AI answer engines like ChatGPT, Perplexity, and Google's AI Overviews. Instead of tracking only blue-link position, it measures whether specific customer questions produce a citation of the client's business, and it builds the structural signals, consistent facts, real reviews, clearly quotable pages, that make a citation more likely. The distinction matters because tactics that move a page up in classic Google results do not automatically make an AI engine choose to cite it.
That is a real shift, not a rebrand. Search engines rank documents. AI engines select sources to summarize into an answer, and the selection criteria overlap with SEO (crawlability, authority, freshness) but are not identical to it (entity clarity, extractable facts, consistent NAP data, quotable structure). An agency that only knows the first skill set is an SEO agency with a new landing page.
Everything else is packaging.
We run a fixed set of real customer questions against ChatGPT, Perplexity, and Google AI Overviews every month, and log whether the client is cited, what is said, and how that changes over time. The exact prompts and answers are re-runnable by the client at any point, not summarized into a vague "visibility score."
Entity clarity, facts that match everywhere (site, GBP, directories, schema), real reviews an engine can verify, and pages written to be quoted in one self-contained sentence. This is slower and less glamorous than a ranking hack, and it is the part that survives algorithm changes because it is not a trick.
We report how long a ranking or a citation held, through updates, not a single best-day screenshot. A term that hits No.1 for a week and drops is a different result from one that holds through several updates, and most reporting does not distinguish between them.
The pitch to be skeptical of is the one promising a ranking or a citation inside ChatGPT on a timeline. AI engines make that call themselves; there is no paid placement or submission form to buy your way past it. What an AI SEO agency can honestly do is measure AI-answer presence with re-runnable prompts, build the structural signals engines draw on, and report the pattern over time. That is a narrower promise than most pitches make, and it is the true one.
A best-day screenshot is not a record. We report how long it held.
One data point, honestly labeled.
| Austin luxury custom home builder | Head terms tracked daily against a named competitor set. Visibility fell with the market during the July core update and recovered afterward. Positions in this category move week to week and we report them as a range, not a snapshot. |
| Ventura County medical clinic | Four tracked terms moved to No.1 in a single four-week cycle, including a flagship term that went from position 8 to 1; 27 of 43 tracked keywords now sit in the top 3; organic sessions up 19% month over month. |
| North Florida aesthetics clinic | Google reviews removed from the profile were caught by our monitoring within the cycle and reinstated in roughly 48 hours. |
Source: Client Studio internal tracking, Semrush Position Tracking and GA4. Client names withheld; anonymized figures available on request.
The builder result is the useful one here. A core update is exactly when thin pages and manipulated signals get flushed out. Terms built on real entity structure, genuine reviews, and consistent facts did not just survive, they pulled ahead of competitors who dropped. That is the case for structural work over the trick of the month.
Most won't have good answers to all five. Ask anyway.
| Ask for | What we show |
| A sample of prompt-level AI tracking | Exact prompts run monthly across ChatGPT, Perplexity, and AI Overviews, with dated results the client can re-run independently. |
| Position durations, not current snapshot | How long each ranking or citation actually held, including whether it survived the July 2026 core update. |
| Source and date on every claim | Every number tied to a named tool (Semrush, GSC, GA4) and a pull date. No "results may vary" language standing in for a citation. |
| Published pricing | Findings, Baseline, and Authority pricing posted openly on /pricing/, not withheld for a sales call. |
| Exclusivity terms | One client per category per market, stated up front. |
Three steps, priced per market.
An external-only scan, no logins or account access required, delivered in about 48 hours. It shows what is currently visible, missing, or broken across search and AI answer surfaces before you commit to anything. Start at /findings/.
$1,500 for the first market ($2,500 in a major metro), $500 for each additional market. The full fee is credited to month one if you continue, and refunded outright if we don't verify at least 3 defects across 4 published classes.
$3,000/mo for one market ($4,500 major metro), plus $1,200 per additional market ($1,800 major metro). Initial term is 90 days, then month-to-month. Every asset built stays yours on exit. One client per category per market. Full detail at /pricing/.
The free Market Findings report is external-only, takes no logins, and lands in about 48 hours. Compare it, prompt for prompt, against whatever your current agency's dashboard is telling you, and see whether the two stories match.
What your buyers ask AI, every day
“best seo agency for [your industry]”
“should I hire an seo agency or do it myself”
“how do I know if my seo agency is working”
We run questions like these monthly, on every engine, and log exactly who gets named. If it is not you, that is the gap we work.
A regular SEO agency optimizes for classic ranking factors and reports traffic and position. An AI SEO agency adds a second practice: monthly prompt-level tracking of whether the business is cited inside ChatGPT, Perplexity, and AI Overviews for real customer questions, plus structural work (entity clarity, consistent facts, quotable pages, verified reviews) aimed at earning those citations. Both should report positions and citations held over time, not one-time traffic spikes.
AI-answer tracking is the monthly practice of running a fixed set of real customer questions against ChatGPT, Perplexity, and AI Overviews and recording whether the business is cited, what is said, and how that changes month to month. There is no public API for "position" inside an AI answer, so the method has to be the exact prompt, the exact response, and a dated log the client can re-run to verify independently.
Pricing should be published, not hidden behind a sales call. The free Market Findings report comes first. A Market Visibility Baseline runs $1,500 for the first market ($2,500 major metro), $500 per additional market, credited in full to month one. The ongoing Market Authority retainer is $3,000/mo for one market ($4,500 major metro) plus $1,200 per additional market ($1,800 major metro), starting with a 90-day term. Full detail on /pricing/.
No, and any agency that guarantees it is not being straight with you. Nobody, including us, can place a specific business inside a specific AI answer on demand. What can be done honestly is measuring current AI-answer presence with re-runnable prompts, building the structural signals engines draw on, and reporting the pattern over time, with sources and dates attached to every claim.