Off-Page, Digital PR and Citations
Build an off-page SEO system around digital PR, linkable assets, brand mentions and AI-cited sources, with measurable outreach rules and clear quality gates.
Off-page SEO is the work performed beyond your own site to earn independent evidence of relevance, expertise, and reputation. It includes editorial coverage, expert contributions, references, reviews, community discussion, accurate mentions, and a backlink —a hyperlink from another site to yours. The aim is not to manufacture authority. It is to give credible third parties a defensible reason to discuss and cite the organization.
Phase: P14, Off-Page, Digital PR and Citations. Stage: C — Build. Timebox: 5–10 working days to build the source map, campaign asset, target list, and measurement baseline; then a 6–12 week outreach and monitoring cycle. Owner: the digital PR or off-page SEO lead is accountable, with subject-matter experts, content, brand, legal, and analytics owners approving claims and measuring results.
This phase separates link acquisition that still creates value from activity whose apparent scale hides weak editorial merit. Digital PR , linkable assets—original resources designed to earn references through their utility or evidence—relationship-led outreach, expert commentary, and broken link building still work when they improve a real publication. Bulk guest posts, interchangeable contributed articles, reciprocal-link schemes, and low-quality directories have been devalued because the placement exists primarily for the link rather than the reader.
Why this phase, and why here
P14 begins after structured data and entity building because outside evidence should reinforce a settled identity. It consumes the approved entity register, canonical pages, aliases, expert records, content inventory, research priorities, competitor gaps, and live assets produced by earlier phases. Those inputs determine who the organization is, which claims it can prove, which page should receive a reference, and which topics matter commercially.
Running outreach before those decisions are stable spreads contradictions. A publisher may use an old product name, link to a page scheduled for consolidation, quote an unsupported claim, or describe two similarly named people as one expert. The resulting coverage can rank, circulate, and feed AI answers long after the site is corrected. External evidence is slower to repair than an owned page.
Skipping P14 leaves the site speaking for itself. Good pages may rank, but the organization lacks independent corroboration and may be absent from the sources an answer engine already trusts. Running it as a classic links-only campaign creates a second blind spot: an unlinked brand mention can still place the brand, category, attributes, and expert claim in a retrievable third-party document. In an AI-citation environment, that context may matter even when no clickable link transfers authority.
The output is not a count of emails or links. It is a documented system that connects priority audience questions to citable assets, credible source targets, honest outreach, verified coverage, mention reclamation, and ongoing measurement.
Inputs and outputs
| Direction | Item | Why it is needed | Acceptance condition |
|---|---|---|---|
| Input | Entity and claims register | Publishers need consistent names, biographies, product facts, and proof. | Every campaign claim has an approved wording, source, owner, and canonical destination. |
| Input | Priority topics and prompts | Outreach must support business-relevant questions, not generic visibility. | Each campaign maps to an approved topic, audience, market, and prompt or search need. |
| Input | Content and competitor gap evidence | Existing coverage reveals what audiences cite and what is genuinely missing. | The brief identifies current winners, missing evidence, and why a new asset improves the record. |
| Input | Live canonical assets | A publisher cannot cite a draft, unstable URL, or gated proof it cannot inspect. | Destination URLs return successfully, are indexable where intended, and expose the promised evidence. |
| Output | Source landscape | The team needs separate lists for classic publications, communities, associations, and AI-cited sources. | Every target has evidence of relevance, target type, priority topic, market, and a named relationship owner. |
| Output | Linkable-asset brief and published asset | Outreach needs something editorially useful enough to reference without a favor. | The asset has a unique evidence base, methods, expert owner, canonical URL, and update rule. |
| Output | Qualified outreach and reclamation register | Activity must remain reviewable and suppress irrelevant or repeated contact. | Every contact records the editorial reason, person, status, next action, evidence, and stop condition. |
| Output | Placement and citation baseline | The next phase needs outcomes tied to sources and landing pages. | Each verified result records URL, date, source type, link or mention status, cited page, target topic, and referral or citation baseline. |
The placement register is the contract with P15. It preserves enough source, URL, date, campaign, and landing-page detail to connect off-page work to later behavior and revenue without guessing.
Ready to Monitor Your AI Visibility?
Track how AI chatbots mention your brand across ChatGPT, Perplexity, and other platforms.
The checklist
Every checklist item carries five fields: what, why, how, tool, and done when. An outreach action is not complete because it was sent; it is complete only when its state and evidence are recorded.
1. Freeze the entity, claim, and destination contract
What: select the exact entity name, spokesperson, approved claims, supporting evidence, and canonical page for the campaign. Why: external inconsistencies are difficult to retract and can teach search or AI systems the wrong identity. How: copy approved records from P13, test each destination, and create publisher-ready wording for names, roles, dates, units, limitations, and methodology. Tool: entity register, claims sheet, CMS, browser, and legal or brand review. Done when: 100% of campaign claims have a source and owner, all destination URLs are live, and zero unresolved naming or identity conflicts remain.
2. Map the source landscape, including sources AI already cites
What: build four target views: editorial publications, specialist or institutional sources, community and review sources, and domains or URLs already cited in tracked AI answers. Why: the sites that pass traditional link metrics are not automatically the sites answer engines retrieve for the audience’s questions. How: review current answers for priority prompts, record cited domains and exact URLs, compare repeated sources across providers, and classify whether each source accepts data, expert comment, tools, corrections, product evidence, or community participation. Do not assume that a cited source is pitchable.
Tool: AmICited Sources, full answer records, search results, backlink data, publication pages, and relationship history. Done when: every priority topic has at least 10 verified relevant sources, the first campaign list contains at least 25 qualified targets, and every target has a source type, cited-topic evidence, contact route, fit reason, and exclusion check.
3. Choose an asset people can cite for a reason
What: commission one linkable asset: original data, a transparent benchmark, a maintained reference, a practical tool, a tested template, or expert analysis that resolves a documented information gap. Why: outreach cannot compensate for an asset that merely repeats available advice. How: state the audience question, unique evidence, collection method, limitations, quotable findings, visual needs, update cadence, and canonical home before production. A product page can support the evidence but should not replace it with sales copy.
Tool: gap analysis, first-party data, research plan, editorial brief, and claims review. Done when: the brief answers “why this source, why now, and why cite us,” all figures reproduce from the underlying evidence, limitations are visible, and the published canonical URL passes content, legal, and technical review.
4. Build a qualified, relationship-led outreach list
What: identify the editor, journalist, researcher, association owner, resource-page maintainer, or community participant responsible for the relevant subject. Why: a relevant publication sent to the wrong person is still spam, and repeated poor contact damages future access. How: read recent work, record the person’s actual beat, note prior contact and preferences, remove conflicts and suppressions, then write a one-sentence editorial reason specific to that target. Tool: source landscape, publication mastheads, relationship database, and suppression list. Done when: every target has a verified person or documented editorial route, a recent relevance example, a unique fit reason, last-contact date, and an allowed next action; generic role lists are zero.
5. Pitch value, not an anchor-text instruction
What: send concise, personalized pitches offering the evidence, expert, correction, or replacement the recipient can use. Why: editors cite material that improves their work, not a sender’s preferred keyword. How: lead with the finding or reader problem, disclose commercial relationships, link to the evidence, offer accessible supporting material, and use no more than one initial message plus two useful follow-ups unless the recipient engages. Never require a followed link or exact anchor.
Tool: approved pitch template, email or PR platform, relationship log, and asset pack. Done when: every sent pitch has a target-specific opening, a verifiable offer, disclosure where required, send date, response state, and next-action date; unsubscribes and declines are suppressed immediately.
6. Reclaim broken links and unlinked mentions selectively
What: find relevant dead references and accurate unlinked mentions, then request the smallest useful correction. Why: reclamation starts from existing editorial intent and can improve the publisher’s page; it fails when every mention becomes a demand. How: for a broken link, confirm the original purpose and offer a close, live replacement. For an unlinked mention, verify the entity and ask for a link only when it helps readers reach evidence, a definition, or the official source. Request factual corrections whether or not a link is added.
Tool: backlink crawler, mention monitoring, browser, redirect history, and outreach register. Done when: every request includes the exact source URL, quoted context, proposed destination or correction, relevance rationale, contact result, and final status; zero unrelated replacement pages are pitched.
7. Treat community and UGC sources as participation, not inventory
What: identify user-generated content —reviews, forum posts, videos, and community discussions—that shapes priority answers. Why: these sources often contain firsthand experience and comparative language that vendor pages cannot credibly supply. How: monitor recurring cited threads, answer openly under a real affiliation, correct material errors with evidence, support genuine customers, and let community moderators decide what remains. Do not seed fabricated questions, votes, reviews, or personas.
Tool: AmICited UGC Influence, community search, support records, and approved participation policy. Done when: priority cited UGC URLs are assigned as monitor, participate, customer-support, correction, or no-action; every participation has a disclosed identity and evidence; fabricated or incentivized undisclosed activity is zero.
8. Verify, classify, and monitor every result
What: confirm live outcomes and classify them as followed link, nofollow or sponsored link, unlinked mention, correction, community reference, or no placement. Why: a publisher promise, forwarded email, or syndicated duplicate is not a verified result, and different outcomes should not be forced into one link count. How: open the live URL, confirm the claim and destination, record link attributes and canonical behavior, detect syndication, save evidence, and monitor whether priority sources begin or stop appearing in AI answers.
Tool: browser, backlink data, AmICited Source & Citation Intelligence, analytics, and placement register. Done when: 100% of claimed outcomes have a live URL or are marked unverified, duplicate syndications share one parent record, incorrect claims have a correction owner, and baseline versus current citation presence is recorded for the supported topic.
Tools in AmICited
Open Source & Citation Intelligence at https://app.amicited.com/sources . Start with the priority prompt set and date window. Review cited domains, then switch to URL view to distinguish a broadly trusted publication from one exceptional page. Drill into a domain’s top pages and prompts, and compare provider columns: a source cited by Perplexity for one prompt is not automatically a cross-platform priority.
Use the output to create an AI-source target tier alongside the classic media and backlink tiers. Record the exact prompt, provider, cited URL, date last seen, relationship opportunity, and whether the source can be influenced ethically. The aim is presence in the evidence ecosystem, not a request that a publisher “get us into ChatGPT.”
Open UGC Influence at https://app.amicited.com/ugc to isolate Reddit, YouTube, forums, and other community pages cited for the tracked prompts. Review the exact thread or video before acting. A high domain total can represent one dominant discussion, and “external” sources also include publishers and company sites, so classify the page rather than the hostname alone.
Export or copy the relevant rows into the source landscape with the observation date. AI answers change, so a target list without its prompt, provider, market, and date cannot be reproduced.
Stay Updated on AI Visibility Trends
Get the latest insights on AI mentions, brand monitoring, and optimization strategies.
Decision rules
These thresholds govern campaign quality and handoff. They are operational gates, not promises that a placement will rank or receive an AI citation.
| Finding | Threshold | Decision | Done when |
|---|---|---|---|
| Unsupported campaign claim | 1 or more | Block campaign | Every claim has approved evidence, wording, owner, and limitation. |
| Qualified launch list | Fewer than 25 targets | Do not launch broad outreach | Add genuinely relevant targets or narrow the campaign; never pad the list. |
| Target relevance | No recent topical example or no specific editorial fit | Exclude target | A source URL and one-sentence fit reason are recorded. |
| Personalization | Any pitch could be sent unchanged to 10 unrelated recipients | Rewrite | The opening refers to the recipient’s real coverage and the offer resolves a specific need. |
| Follow-up limit | More than 2 unanswered follow-ups | Stop contact | The record is closed or paused until the recipient re-engages. |
| Unsubscribe, explicit decline, or do-not-contact request | 1 event | Suppress immediately | All active sequences stop and the suppression is shared. |
| Paid or reciprocal ranking placement | 1 proposed placement | Reject | No money, product, or reciprocal link is exchanged for undisclosed ranking credit. |
| Bulk guest-post or low-quality directory source | Any source whose primary value is placement rather than audience utility | Exclude | The source is removed and the exclusion reason is recorded. |
| Result verification | Below 100% of claimed placements | Do not report as earned | Each result has a live URL and classification, or is labeled unverified. |
| Factual error in earned coverage | 1 material error | Open correction within 1 business day | The publisher is contacted with evidence and the outcome is recorded. |
| AI-source evidence | Missing prompt, provider, market, or observation date | Do not prioritize from citation data | All four dimensions accompany the cited URL. |
| Campaign review cadence | No review for 30 days during an active cycle | Pause new volume | Outcomes, suppressions, source movement, and message quality are reviewed first. |
A response rate alone is not a quality target. Ten replies rejecting an irrelevant pitch do not outperform one editor using sound evidence. Track delivery and response for diagnosis, but judge the campaign on qualified relationships, accurate coverage, useful citations, supported topics, referral behavior, and eventually business outcomes.
Deliverable
Hand over one versioned off-page evidence register with linked asset and pitch materials. A spreadsheet, database, or project table is acceptable if it is filterable and exportable without losing the evidence trail.
CAMPAIGN CONTRACT
Campaign | Audience | Market | Priority topic and prompts
Entity | Approved claim | Evidence | Canonical destination | Owner
SOURCE LANDSCAPE
Domain | Exact URL | Source type | AI provider | Prompt | Observed date
Editorial fit | Contact | Relationship owner | Priority | Exclusion reason
OUTREACH AND RECLAMATION
Target | Specific reason | Offer | Sent date | Follow-ups
Response | Next action | Stop condition | Disclosure | Evidence
PLACEMENT AND CITATION LOG
Live URL | Publication date | Parent or syndication record | Source type
Brand mention | Destination | Link attribute | Supported topic | Accuracy status
Referral baseline | AI citation baseline | Last verified | Correction owner
The handoff is accepted when another owner can reproduce why every target was chosen, distinguish a link from a mention or citation, open proof for every reported result, and join the destination URL and campaign ID to analytics in P15.
What goes wrong
The team buys volume disguised as outreach. A vendor reports hundreds of placements across sites with no real audience or editorial standards. Remove placement quotas from the brief and audit the source, page, disclosure, traffic relevance, and editorial reason.
A “data study” has no reproducible method. A striking number earns attention but nobody can explain the sample, exclusions, or calculation. Withdraw the claim, publish the method and limitations, and make the underlying evidence reviewable.
Classic link metrics become the only target filter. The campaign ignores specialist databases, associations, communities, and pages repeatedly cited by AI answers. Maintain separate target tiers and compare them by the audience question they answer.
Every unlinked mention gets an aggressive email. Positive coverage becomes a strained relationship. Ask only when the link materially helps the reader; otherwise thank, monitor, and preserve the accurate reference.
A cited source is treated as controllable. The team pressures moderators, hides affiliation, or tries to place manufactured praise. Use transparent participation and accept no-action decisions.
The pitch asks for exact-match anchor text. The request reveals that the link, not the reader, is the purpose. Let the publisher choose natural wording and classify the result honestly.
Syndication is counted as independent wins. One wire story appears on 40 domains and becomes “40 placements.” Record one parent placement, attach syndicated children, and report both without confusing replication with separate editorial endorsement.
Coverage links to the wrong URL. A campaign wins attention for a temporary page, redirect chain, or retired product name. Freeze destinations before launch and correct material errors quickly.
The team celebrates links without measuring the supported topic. Referral visits, rankings, brand mentions, and AI citations are not automatically caused by one placement, but the campaign still needs dated baselines and annotations so P15 can evaluate plausible contribution.
Next phase
P15 is conversion and revenue tracking . It receives the campaign contract, source landscape, verified placement log, destination URLs, campaign IDs, publication dates, and baseline citation observations. Its owner connects those records to referral sessions, assisted journeys, qualified actions, pipeline, and revenue where consent and data quality allow.
The P14 owner also hands over known measurement limitations: publishers that strip campaign parameters, dark or untrackable referrals, syndicated URLs, nofollow links, unlinked mentions, cross-device gaps, and AI answers that cite a source without sending a visit. P15 must not convert absence of a click into absence of influence, nor claim revenue causality from temporal coincidence.
Handoff is complete when every verified result has a stable source URL, observed or publication date, destination or unlinked status, supported topic, source type, campaign identifier, and accuracy state. The next phase waits if placement totals cannot be reconciled to live evidence.
FAQ
Frequently asked questions
Do unlinked brand mentions help if they do not pass link equity?
How many outreach emails should we send?
Should we pay for guest posts or directory links?
Should every AI-cited source become an outreach target?
When is the phase complete?
More tutorials in this section
Ready to put it into practice?
Free check · 7-day trial · no credit card