Services · SEO and AI search
Ranking was the goal. Being cited is the goal now.
Most searches end on the results page, and a growing share of buyers ask an assistant instead of a search box. Your site either gets extracted and named in that answer, or it does not. We build the site so it can be, and we run a scored probe set that says whether it is.
No ranking promises on this page. Read it and check.
- 01Who fixes a Google Ads account that an AI tool built?
- Youcited
- 02What does a Google Ads audit cost for a small company?
- Two competitorsnot cited
- 03Is offline conversion import worth it for a service business?
- You, one competitorcited
- 04Best local agency for call tracking setup
- A directorynot cited
Score this run · one mark per question
A real set is longer and runs on a schedule. The score moves for reasons we can name.
What changed
The results page answers the question now. The click is optional.
For a long time the deal was simple. Rank, get the click, get the call. The results page now answers many questions itself, with an AI Overview at the top and an extracted snippet under it, and a growing share of buyers skip the search box for an assistant. In both cases the buyer sees whoever the machine could extract and attribute, and never sees the rest of the page.
So a page-one ranking that never gets cited is worth less than it was, and a page the assistant can quote is worth more. The work below is organized around that second thing. Rankings still matter, because citation draws from what ranks. They are an input. The score is citation.
- ~68%
- of searches end with no click
- ~73%
- of B2B buyers consult AI assistants
Rounded third-party industry figure. We did not measure it; the probe set below is what we measure.
Rounded third-party survey figure. Same caveat. Our own reading is the probe score below.
The work
Six packages. Each one leaves a document behind.
Not every site needs all six. The triage call decides which, and the scope is written down with a fixed price before anything starts.
01
Architecture by demand
Every page gets mapped to the searches it exists to answer, and every search you should own gets a page. Pages with no demand behind them get merged or retired. This is where most of the movement comes from, and it is done before any code is touched.
Leaves behindKeyword-to-page map, versioned
02
Technical SEO, on your stack
Rendering, canonicals, sitemaps, redirects, page speed, internal links. On Next.js this is code in git; on WordPress it is theme and plugin configuration we document. Either way the fix is inspectable afterward.
Leaves behindTechnical defect list with fix dates
03
Schema and extraction blocks
Structured data for the entity, the services, and the questions, plus plain-text fact blocks on every page written so an assistant can lift the answer and attribute it. This page has one at the bottom; so does every indexed page on this site.
Leaves behindMachine-readable layer on every page
04
The SOMA probe set
A versioned list of buyer questions, run on a schedule against AI Overviews and the major assistants, scored on whether you were cited and who was cited instead. It is the instrument for everything above.
Leaves behindDated probe scores, run over run
05
Google Business Profile
Categories, services, attributes, posts, and the review response cadence, kept consistent with the site so the local pack and the entity graph agree about who you are.
Leaves behindProfile reconciled to the site
06
Content, only where demand is proven
A page gets written when the map shows a search you should answer and do not. It does not get written to hit a monthly quota. Most owner-led sites need fewer pages than they have.
Leaves behindPages tied to a mapped search
The instrument
How do you know if an assistant cites you? You ask it, on a schedule, and write down the answer.
The probe set starts with your buyers' questions, not keywords. We take the questions that show up in your calls, your search terms, and your sales conversations, phrase them the way a buyer would type them, and version the list. Then each question is run against AI Overviews and the major assistants, and every answer is recorded: were you named, who else was, what did the answer get wrong about you.
The run repeats. When a score moves, the change log says what shipped between the two runs, so movement gets a cause instead of a guess. When the answer is wrong about you, the fact block on the relevant page is the fix.
One probe row · what gets recorded
- Question
- The buyer's phrasing, versioned
- Surface
- AI Overview or a named assistant
- Cited
- Yes or no, verbatim answer archived
- Named instead
- Who the answer chose, and why it could
- Errors
- What the answer got wrong about you
- Run date
- ISO date, every run, no exceptions
Same grammar as the proof ledger: a claim, an instrument, a date.
What we will not sell you
Four things other SEO shops sell that we decline to.
- A ranking guarantee
- Nobody controls the ranking. A guarantee describes the refund terms, not the method.
- A monthly blog quota
- Volume without mapped demand produces pages nobody searched for and dilutes the ones they did.
- Link buying
- It works until it is detected, and then it is your domain that carries the penalty, not the vendor's.
- A ranking report as the deliverable
- Positions are an input. The deliverable is the probe score, the search console read, and the calls that trace back.
Where the organic work stops
Organic pages carry the evidence layer and the questions. Paid traffic lands on separate, noindexed pages matched to the ad that sent it. Keeping those two apart is a rule, not a preference, and it is why the landing page work is its own service.
Landing pages and websites ›Where the calls get counted
A probe score and a search console read are readings of the top of the funnel. Whether the visit became a call is a tracking question, and that is Signals Foundation.
On the record
- What does MyWebReps do for SEO?
- Site architecture by demand (each page mapped to the searches it should answer), technical SEO on Next.js or WordPress, schema and extraction blocks so AI assistants can cite the site, Google Business Profile upkeep, and content only where demand is proven.
- What is AI-search visibility?
- Whether AI Overviews and AI assistants cite your company when a buyer asks them a question you should be the answer to. Industry data puts roughly 68% of searches at zero clicks and roughly 73% of B2B buyers consulting AI assistants, so a ranking that never gets cited is a smaller asset than it used to be.
- What is the SOMA probe set?
- A versioned list of the questions your buyers actually ask, run on a schedule against AI Overviews and AI assistants. Each run records whether you were cited, who was cited instead, and the date. The score moves for reasons we can name.
- Does MyWebReps guarantee rankings?
- No. Nobody controls the ranking, and a shop that guarantees one is describing its refund policy, not its method. We publish what we changed, when, and what the probe set and the search console read afterward.
- Do I need new content?
- Only where demand is proven. Most owner-led sites need fewer, better pages, each mapped to a real search, before they need a blog. We will say so if that is the case.
- Next.js or WordPress?
- Both. The technical work differs (rendering, schema, sitemaps, redirects), the architecture work does not. We work in whichever your site already runs on, and we will tell you when a rebuild would pay for itself.
- What does SEO and AI search work cost?
- A fixed fee for a defined scope, published before work starts. Ongoing probe tracking and upkeep run monthly. The number is on the triage call, not hidden behind a form.
- How do I start?
- Book a 15-minute triage call. Search console access is optional and gets you a live read on the call.