Make ten existing pages easier to quote, inspect, and cite

The fixed deliverable

A content director or head of SEO at an SMB or mid-market company pays $1,500 once for a citation-readiness retrofit of ten existing public pages on one domain. The buyer receives a page inventory, a claim-to-source matrix, ten annotated rewrite packets, one machine-readable content contract, a frozen twenty-prompt observation manifest, a baseline worksheet, and one post-publication delta report. The rewrite bundle arrives within ten business days after scope, payment, and the public URL list are accepted. If the buyer publishes the accepted changes, the delta report arrives within five business days after the thirtieth calendar day following buyer-confirmed publication. There is no renewal or automatic recurring charge.

Acceptance boundary. The acceptance test is the delivered evidence bundle, never whether an answer engine chooses to cite a page. A page can pass the declared content checks and still receive zero observed citations. That is a valid result, not a reason to relabel the measurement or weaken the rules.

What “citation ready” means here

Citation readiness is a reviewable page property, not a prediction about an opaque external system. A ready page states the question it answers, keeps the answer close to the question, distinguishes sourced facts from the publisher's own position, links material factual claims to inspectable sources, uses headings that describe the section beneath them, exposes update context where freshness matters, and avoids hiding the core answer behind an interaction. Those choices can make a page easier for humans and machines to inspect. They do not compel an answer engine to retrieve, trust, quote, rank, or cite it.

The retrofit therefore uses three separate ledgers. The content ledger records what changed and why. The source ledger binds each material factual claim to a source URL, publisher, retrieval date, and interpretation limit. The observation ledger records what a named answer surface returned for each frozen prompt. A source link is not an observed citation; an observed URL string is not provider-authenticated grounding; a crawler fetch is not a referral; and a referral is not a sale. Keeping those event classes separate is part of the product.

1. Freeze scope. The buyer and reviewer name ten public URLs, the audience, one business objective, and twenty prompts before results are read.
2. Inventory the page. Each claim, heading, source, date, answer block, internal link, and structured field receives a deterministic status: present, absent, invalid, or unknown.
3. Draft the retrofit. Proposed copy stays in annotated Markdown. Every changed factual claim carries its source or is explicitly marked for buyer substantiation.
4. Run the content gate. A failed link, mismatched heading, unsupported fact, missing denominator, or unevaluable rule remains visible. The gate is never trained to approve its own output.
5. Buyer publishes or declines. Production access stays with the buyer. A declined rewrite remains a delivered proposal and is excluded from the published-page denominator.
6. Remeasure once. Only buyer-confirmed published pages are observed on the same prompt and answer-surface protocol after thirty days. The report shows planned, run, readable, unknown, excluded, and cited counts.

The measurement does not move after the answer arrives

The twenty prompts are written from the accepted page questions and frozen before the baseline. Each row names the prompt text, target URL, target domain, locale, named answer surface, session state, search or grounding state when knowable, run date, response hash, exact answer-text URL tokens, and provider citation targets when the interface exposes them. Missing metadata, an empty response, a blocked page, a provider error, or an unreadable result is UNKNOWN. It is never counted as a miss, a zero, or a pass.

The numerator is equally narrow. A provider-authenticated citation requires a target in the provider's structured citation or grounding metadata. An exact target URL printed in prose is reported separately as an answer-text URL mention. A brand or page-name mention without the target is separate again. The final comparison shows counts as k of n readable observations, with the planned denominator beside it. It does not collapse different providers into a “visibility score,” infer sentiment, choose favorable screenshots, or replace missing rows with an average.

No causal claim. A before/after difference is an observation across two dated panels. Answer engines, indexes, competitors, prompts, model revisions, and web sources can change during the interval. The report records the retrofit and the result but does not claim the retrofit alone caused a gain or loss.

Owned proof: the method is willing to print zero

Our public Cited by AI? observation page is the operating proof for the measurement boundary. Its sealed dataset-targeted panel contains 6 prompts on 2026-08-09. The source reader counted 4 readable misses, 2 exact asset-name mentions without a target URL, 0 exact target URL mentions, 0 unknown rows, and 0 provider-authenticated citations.

That deliberately unimpressive result matters. A measurement system that can only publish a positive score is marketing copy, not evidence. The panel used one local, ungrounded control assistant on one date, so it says nothing about a commercial provider and predicts nothing about a client retrofit. It proves a narrower capability: the manifest, hash, label hierarchy, unknown handling, and least-flattering publication rule can preserve a zero-citation result without promoting crawler logs or name mentions into citations.

Counted market and owned-query receipt

The official vendor sources below returned HTTP 200 on 9 August 2026 in Arizona. They show that content and SEO teams can buy prompt-based AI-search monitoring. They do not show that software and this service are substitutes, establish our price, or prove demand. Prices and plans can change; the machine-readable receipt keeps each fetched URL, response hash, observation, and interpretation limit.

Search Console language already reaching the owned property

The Search Analytics API returned 33,663 query-dimension rows for 2026-07-10 through 2026-08-08. The fixed language filters produced the counts below. Query rows can overlap between filters, so they are not summed into a market total.

The zero-click counts stay visible. Impressions prove only that at least one owned page appeared for a recorded query. They do not identify a buyer, establish keyword volume beyond this property, or count a referral. This page creates a specific demand instrument: Search Console can count impressions and clicks to this URL, while the product-specific request path can count a scope request separately. Our fetches and crawlers remain validation traffic.

What happens after a scope request

1. Coarse fit check. The buyer names the public domain, approximate eligible page count, content type, audience, and answer questions. The request form accepts no file or account access.
2. Written handling plan. If the fit is accepted, the human parties define the ten public URLs, rights, excluded topics, transfer method, source responsibilities, prompt protocol, deliverables, price, and dates before payment.
3. Baseline and retrofit. The frozen baseline runs first. The buyer then receives the claim/source matrix, annotated rewrites, deterministic audit, schema suggestions, and handoff instructions.
4. Buyer-controlled publication. The buyer reviews, changes, publishes, or rejects each proposal. We do not receive production credentials. Only confirmed published pages enter remeasurement.
5. One dated follow-up. Thirty days after confirmation, the same protocol runs once. The report shows the complete denominator, unknowns, exclusions, exact URL mentions, provider citation targets, and per-page changes without a causal or ranking claim.

Request the ten-page fit check

Questions to settle before buying

Will you guarantee that an assistant cites our pages?

No. The offer delivers a reviewable retrofit bundle and a fixed before/after observation. External answer systems remain outside our control.

Do you write ten new articles?

No. This pilot starts with ten existing public pages and returns annotated retrofit packets. It is not the blog lane, a publishing retainer, or a bulk content factory.

Can you make unsupported marketing claims sound authoritative?

No. A buyer assertion without acceptable substantiation stays labeled as buyer-provided, unsupported, or excluded. The service will not invent a source or disguise opinion as independent fact.

Do we need a monitoring subscription?

No subscription is included or required by this offer. The written scope identifies suitable answer surfaces and who lawfully operates them. If a paid vendor is required, the buyer supplies it under a separate decision; this public test spends nothing.

What if we do not publish the rewrites?

The retrofit bundle is still delivered, but no page is silently counted as treated and no post-publication citation delta is claimed. The report states that remeasurement was not supportable.

What is the external discovery lane?

Content and SEO teams already evaluate AI-search work through the Peec AI agency directory and comparable GEO tool ecosystems. A directory application is not claimed here: qualification and any partner terms must be checked before a truthful listing. The indexable page and its request instrument are the unblocked first step.

This is an offer test. No customer, payment, ranking gain, citation lift, referral, partnership, or product-market fit is claimed.