Public data index › Citation Repair & Source Integrity Sprint
Know which citations broke before your readers do
For $349 once, get a dated review of up to 50 public articles and 250 distinct external citations, plus 20 prioritized claim-to-source checks and a specific, reviewable edit list.
This sprint is for the moment after a domain or CMS migration, an editorial handoff, an acquisition or rebrand, or a cleanup pass over AI-assisted drafts. In each case, article text can survive while links quietly stop meaning what they meant. Broken and inconclusive sources stay separate, so an access barrier does not become an instruction to delete a good citation.
Request a scope review
$349 one time
Delivery is targeted within five business days after written scope acceptance and receipt of the public URL set. The form is not checkout: it charges nothing, starts no work, and commits you to nothing.
No credentials, files, pasted articles, private URLs, or CMS access are requested.
What you receive
The deliverable is an inventory and recommendation set. Your team makes and publishes every edit. Three artifacts arrive together:
1. Dated citation inventory
One row per citation occurrence, with the source article, anchor text and surrounding claim context, exact requested URL, final URL after redirects, HTTP or transport observation, closed status, and checked timestamp. A repeated external URL is checked once against the 250-URL cap but remains listed against every article where an editor needs to find it.
2. Bounded review notes
Every non-LIVE row receives a concise human triage note explaining what was observed and why the status was assigned. The note describes the observation, not a theory about the source publisher. It will not turn a rate limit, paywall, or bot gate into a claim that the cited material disappeared.
3. Up to 20 repair recommendations
For each selected claim-to-source review, you receive a primary-source replacement suggestion where one can be established, or an explicit recommendation to remove the citation, narrow the prose to what a surviving source supports, or leave the item unverified for your own review. If you later provide authorized, non-sensitive HTML or Markdown snippets, the instructions can be formatted as patch-ready find/replace steps. Otherwise they identify the public page, anchor, and proposed action.
One recheck is included after your team applies changes. Request it within 30 days of delivery; it covers the citations flagged in the original run and returns an updated dated record.
Four statuses, with uncertainty preserved
Every citation lands in exactly one closed status. The definitions are public so an editor can audit any row against the rule that produced it.
LIVE
The exact URL returns a 2xx response, and its body was inspected and does not appear to be a not-found page, parked page, login or consent interstitial, or unrelated destination. A 200 response by itself does not earn this label.
MOVED/REVIEW
The final URL differs materially from the requested URL, or the destination may no longer contain the cited resource. Routine HTTP-to-HTTPS upgrades, host canonicalization such as adding www, and trailing-slash normalization do not trigger this status by themselves.
ROTTED
A confirmed 404 or 410, or a redirect whose destination clearly lost the cited resource after body inspection. This is the only status that asserts a citation is broken.
UNVERIFIED
A 401, 403, 429, or 5xx response; timeout; paywall; bot gate; consent barrier; or any other inconclusive response. A source we cannot verify stays UNVERIFIED; it is never relabeled as broken or silently carried forward as current.
No deletion advice from an inconclusive fetch. No removal or replacement recommendation is based only on a single-vantage transport result. An inconclusive source goes to human review where it is publicly openable, or back to the customer where it is not. A replacement finding means only that the cited assertion appears in that source at check time. It does not certify that the assertion is true, current, legally sufficient, or best available.
How the 20 deeper reviews are chosen
The priority order is stated before the work starts. Buyer-designated pages and claims consume review slots first, in the order you name. This lets you put sales-critical, compliance-sensitive, or handoff-critical pages ahead of low-traffic archive material. Remaining slots go first to confirmed ROTTED citations, then to MOVED/REVIEW rows where the destination appears materially different.
If your priority list is shorter than 20 items, the remainder is filled by that second rule. The delivery names every citation reviewed in depth and distinguishes it from citations that received inventory and status review only. There is no hidden sample presented as full coverage.
What this catches
- External sources returning confirmed 404 or 410 responses after a migration.
- Redirects that land on a generic hub, category page, or homepage instead of the cited document.
- Soft-404 and parked-domain pages that return 200 and look live to a status-only checker.
- Login, consent, or access interstitials that a crawler might count as a successful fetch.
- Duplicated and inconsistent citation targets after an editorial handoff.
- AI-assisted drafts that cite a plausible URL resolving to unrelated material or nothing.
What it does not claim to catch
- Whether every accessible source supports its attached claim; that deeper check is limited to the 20 selected reviews.
- Whether any factual assertion is correct, current, complete, or suitable for a regulated use.
- Content behind authentication, paywalls, bot gates, or other access controls. We do not bypass them.
- Anything about rankings, indexing, crawl budget, traffic, leads, or revenue.
Method and owned proof
Our external checker records the exact URL as published, the final URL after redirects, the result or error, and the check date. Broken and unverified are separate first-class outcomes, not a filter applied after a crawler has already guessed.
Counted on 2026-08-09: our internal-link workflow scanned 11,335 Markdown files and surfaced 113 findings across 77 files for triage. Those are internal-link workflow counts. They show that a bounded inventory-and-triage process runs on real files; they are not external-citation results, customer outcomes, evidence of market demand, or a universal accuracy claim. This offer is a demand test and has no purchase history to cite.
Google Search Central's link guidance explains what a machine can follow and why descriptive anchor context matters. RFC 9110 defines HTTP semantics, including the difference between a missing resource, a redirect, and a server refusing to answer. These sources explain method boundaries. They are not SEO promises.
Who this fits
- A content manager, SEO lead, or agency editor who owns a migration or corpus handoff.
- A B2B knowledge base with roughly 20 to 50 public articles and meaningful external sourcing.
- A team cleaning AI-assisted drafts before an editorial or acquisition review.
- A buyer who can apply changes and wants a defensible record of what was observed and when.
Who should not request it
- Anyone who needs us to log in, change a CMS, or publish edits.
- Anyone seeking a full-corpus fact check or a legal, medical, or financial accuracy opinion.
- Anyone needing an affidavit, evidentiary certification, accessibility audit, or guaranteed search outcome.
- Anyone submitting private pages or a domain they do not own or control.
The exact workflow
- Scope request. You submit only public-domain information and counts. No payment is taken.
- Suitability check. We confirm the domain is public, controlled by the requester, within the caps, and workable. An unsuitable request is declined before payment.
- Written scope. We state article count, citation ceiling, priority list, start condition, and target delivery date. Work starts only after you accept it.
- Inventory. External citations are extracted with anchor and claim context, then deduplicated for transport checking.
- Check and body inspection. Each URL is requested, redirects and response body are examined, and the observation is timestamped.
- Human triage. Every non-LIVE row is reviewed before its status closes. Inconclusive results remain UNVERIFIED.
- Claim review. Up to 20 items are reviewed in the disclosed priority order.
- Delivery and recheck. You receive the files, apply any edits, then may request one recheck within 30 days.
Data boundary
The public request form asks for a company or work email, one customer-controlled public domain, a public sitemap or public article-URL list, approximate counts, your trigger and deadline, and buyer-selected priority pages or claims. Do not paste article text, internal notes, customer information, credentials, tokens, or secrets into the form.
We do not accept CMS or production access, private or staging URLs, source files, source code, personal data, or credentials. We do not bypass paywalls, bot gates, consent barriers, or authentication. Optional short snippets for find/replace formatting are accepted only after written scope and only when authorized and non-sensitive. You remain responsible for every edit and publication decision.
Common questions
Why not run a free broken-link crawler?
You should run one; hard 404 discovery is useful and available at no cost. The paid work is response-body inspection for soft failures, preserving 403/429 uncertainty, attaching claim context, human triage, and the 20 bounded source-repair decisions an editor can review.
What if the corpus is over the caps?
We either scope to the 50 pages you prioritize or decline. We do not silently sample a larger corpus and describe partial coverage as complete.
Will you fix the pages?
No. The instructions are designed for your editor to apply. Credentials and publishing authority stay on your side.
Can this promise search performance?
No. The output is a dated citation-state record and a reviewable edit list, not a ranking, indexing, traffic, accessibility, or revenue service.
Request the sprint
Send the public domain or sitemap, estimated article and citation counts, the event that triggered review, your deadline, and the pages or claims you want prioritized. Do not send article text, internal documents, credentials, private URLs, or secrets.
$349 once after written scope acceptance. The request form takes no payment.