Find Link Building Opportunities with Claude or ChatGPT

A list of websites that link to your competitors is a research starting point. It does not tell you which editors would find your work useful, which pages they still maintain, or whether you have anything worth suggesting.

The useful output of link prospecting is a short list of specific pages with a defensible reason to contact their owners. For each one, you should be able to finish this sentence:

This page serves readers who need ___, and our resource helps them by ___.

If you cannot fill in both blanks without mentioning your own rankings, keep researching. Claude or ChatGPT connected to RankParse can organize the evidence and produce a shortlist for review. Your job is to check the editorial opportunity before anyone makes contact.

Start with something another publisher can use

Choose the destination page before you search for prospects. Otherwise, you will collect websites and invent reasons to approach them later.

A useful destination might be an original dataset with a transparent method, an accessible template, an expert explanation of a difficult process, or a calculator whose assumptions readers can inspect. A commercial page can also deserve a link when it directly answers a purchasing or sourcing question. The requirement is usefulness in context, not a particular content format.

Write a small asset brief:

• Reader: Who needs this resource, and in what situation?
• Task: What does it help that person do?
• Evidence: What supports the advice, figures, or calculations?
• Difference: What can readers get here that existing resources leave out?
• Friction: Can someone use the resource without registering, paying, or navigating a sales funnel?

Friction is not automatically disqualifying. A paid industry report can still be valuable. But an editor needs to know what readers will encounter, and an openly available worksheet may fit a resource list better than a gated ebook.

If your asset has no clear advantage over the pages already cited, improving it may be the best first prospecting task.

Set up a RankParse research packet

Connect Claude or ChatGPT through the secure OAuth route in the MCP connection guide. Enable the connector in the conversation where you will do the research. This route can read backlink evidence and check public pages; contact discovery and sending are outside its toolset. No API key belongs in the web connector URL.

Give the assistant an asset brief and a small research packet. The following packet uses fictional domains; replace them before running anything:

Our domain: hallwise.example
Asset URL: https://hallwise.example/resources/maintenance-worksheet
Reader: volunteer managers of community halls in England
Task: schedule routine maintenance and identify professional checks
Difference: editable schedule, role assignments, guidance boundaries
Access: free download, no registration
Confirmed topic competitor: hallplanning.example
Relevant competitor resource: /guides/hall-maintenance
Exclude: paid placement sellers, investor-only publications
Existing relationships: [attach publisher list and contact history]
Review capacity: five publisher pages this week
Output: evidence-backed shortlist, not emails

An existing relationship list matters even when a domain appears in the gap. RankParse may not have observed a recent link, and an outreach owner may already be speaking to that publisher. The worksheet's jurisdiction and access requirements help the assistant reject superficially similar pages before you spend time reading them.

Stage 1: find domains, preserve the scope

Start with get_competitor_gap. Its domain is your domain and vs is the competitor, both without URL paths. Reversing them answers the opposite question. Use a small limit for a reviewable output, then inspect relevance before requesting page details.

Use my packet. Call get_competitor_gap with domain=[our domain],
vs=[confirmed competitor domain], limit=15. Preserve from_domain,
total_links and from_domain_score plus partial, scored, returned,
total and crawl_release from the response.

Explain the coverage limits. If scored=false, treat the scores as
unknown. Select at most five domains for page-level research, with
one tentative audience-fit reason each. Domain names alone are not
proof of fit; label those judgments hypotheses. Stop at this table.

The competitor-gap documentation explains the distinction between returned and total: the first counts delivered rows; the second counts gap domains found and is a lower bound when partial is true. Large competitor profiles are sampled before scoring. Even a scored list is neither a complete market map nor an editorial recommendation.

If you are unsure which competitor asset matters, use get_top_pages with that competitor's domain and a modest limit. Its url, inbound_links, and referring_domains help you identify linked resources to inspect. A heavily linked homepage may reflect brand recognition; it does not establish demand for your worksheet. Do this discovery only when the asset relationship is unclear.

These prompts scope the assistant's work, but they are not enforced spending limits. Review actual tool calls and use account controls when a hard spending boundary is needed. check_credits can show the available balance before the research stage.

Stage 2: turn domain candidates into referring pages

For each selected publisher, query the competitor, with the publisher as from_domain. This retrieves the citation you want to understand. Use sort=recent to request URL-level evidence:

For the selected publishers, call get_backlinks with:
domain=[competitor domain], from_domain=[publisher domain],
sort=recent, link_type=hyperlink, limit=10.

Keep from_url, to_url, anchor_text, rel and crawled_at. Group rows by
publisher and competitor destination. Prioritize citations relevant
to our asset; do not treat a homepage mention as a resource citation.
Mark absent evidence as unknown. Return at most five exact referring
pages with a hypothesis about why each cited the competitor.

The backlinks endpoint can return aggregated domain rows under sort=importance: their from_url is a homepage, link_type is aggregate, and the crawl date is empty. Those rows cannot identify a paragraph to pitch. If the assistant presents them as editorial pages, ask it to discard that interpretation and retrieve URL-level rows with sort=recent.

from_domain is an exact filter. A mismatch between a publisher's main domain and the domain represented in the evidence can explain an empty response. Preserve the actual returned domain when narrowing. There is no offset pagination; repeatedly asking for “the next page” does not uncover an exhaustive list. Also, total on a backlinks response counts the rows returned, unlike the gap endpoint's total.

Stage 3: check the opening and produce a usable row

Use get_page_seo on shortlisted from_url values and your proposed destination. Record data.status_code, data.final_url, data.title.text, and relevant data.headings as live page checks. An API response can succeed while the target's status_code is 404. A heading match does not prove the passage or citation is still there.

Open the referring page in a browser, or supply its current passage from a separate browsing tool. Confirm the exact citation, intended reader, competing resources, and submission policy. get_page_seo returns internal and external link counts, not a destination list or the surrounding paragraph. It cannot make this editorial check for you.

Here is an illustrative transformation, not a RankParse run. The fictional values show the evidence trail rather than an expected result:

Now ask the assistant to reconcile these sources:

Here are my page-review notes, source URLs and check date: [paste].
Revise the candidate table. Keep the snapshot crawl date separate
from my review date. For each row, quote or summarize the checked
passage supporting fit and identify anything that contradicts it.

Use Ready, Research, Improve asset or Exclude. Ready requires a
specific reader benefit and a checked submission route. Export the
reviewed rows as a Markdown table I can copy into our tracker.
Do not fill missing contact details or claim you checked my notes.

The resulting illustrative tracker row would read: Community Rooms volunteer resources; Ready; offer an editable routine-check schedule alongside the existing responsibilities guide; submission form checked by the reviewer; owner: Maya; next action: prepare a resource suggestion. The evidence columns retain both the backlink crawl date and the human review date. If the existing guide already includes an equivalent schedule, change the decision to Exclude or Improve asset rather than manufacture a gap.

Repair weak results before collecting more

For a solo researcher, one competitor and five inspected pages can fill the next work session. A larger team can run separate competitor packets for different assets, but should combine the results into one publisher register before contact. Keep asset-specific fit reasons: one publisher may suit your maintenance worksheet and reject your booking software. Deduplication should prevent duplicate outreach without erasing those different editorial needs.

Follow the reason for an existing link

Competitor backlinks can reveal where your subject receives editorial attention. The key is to inspect why a link exists.

Begin with competitors for the resource's topic, including publishers that do not sell your product. A facilities software company might compete in search with an industry association, a university department, and a specialist consultant. Your SEO competitor analysis should distinguish those search competitors from companies your sales team encounters.

Look at the pages attracting relevant links, then open several referring pages. Classify the reason for each citation:

This prevents a common mistake: treating every competitor link as an opportunity you can reproduce. A founder interview, sponsorship, historical partnership, or customer integration may have no equivalent for your business.

Also inspect the current page. A backlink export can describe a page that has changed, moved, or disappeared. Search engines and link tools have different coverage and update schedules; an export is evidence of discovery, not proof of today's page state.

Search for pages with a specific editorial opening

Use several discovery routes, because each produces a different kind of opportunity.

Resource lists that match the reader's task

Search the web for combinations of the audience, task, and resource type. For example:

community hall maintenance checklist resources
"volunteer facilities manager" "templates"
"building maintenance" "resource library"

Quotation marks narrow the phrasing and may miss useful variations. Broaden the search when results become repetitive.

Open the pages before recording them. A resource library for professional engineers may have little use for a basic volunteer checklist, even when the topic keywords overlap. Read its introduction, existing selections, and submission guidance to understand the intended audience.

Unlinked references to your original work

Search for distinctive wording from a named study, framework, or resource you published. A publisher may refer to your work without giving readers an easy route to the source.

Confirm that the reference really concerns your work. A generic phrase or similar company name is not sufficient. When attribution is clear, a polite request to make the source accessible can be reasonable. The publisher still decides whether a link belongs there.

Missing or inaccessible sources

When a relevant article cites a resource that no longer works, check what the source originally covered before proposing a replacement. An archived description can help establish the subject, but it does not prove that your page is equivalent.

A dead link to a detailed statutory reference is not an opening for a generic blog post. Suggest your resource only if it serves the original purpose, and disclose meaningful differences. Sometimes the useful contribution is simply alerting the editor to the broken reference.

Real relationships and eligibility

Industry associations, customers, suppliers, professional groups, and community organizations sometimes maintain useful directories or educational collections. Check whether your organization has a legitimate relationship or qualifies for inclusion.

Do not manufacture a partnership or make links a condition of a relationship. Follow the publication's criteria, and keep paid placements clearly disclosed and appropriately qualified.

Turn discoveries into decisions: a fictional example

Imagine a fictional company, Hallwise, has published a free maintenance planning worksheet for volunteer-run community halls. The worksheet separates routine checks from work that needs a qualified professional, and explains how to adapt it to local requirements.

Its researcher finds these four pages. The candidates and judgments below are illustrative, not observed campaign results.

The shortlist contains two pages, not four domains. The association has the clearest immediate case. The local guide needs another piece of research before contact. Neither decision requires a minimum authority score.

Notice that the best next action is sometimes internal. If the support service works in another jurisdiction, Hallwise may need a properly reviewed regional version before its worksheet is suitable.

Prioritize with reasons you can inspect

A numeric score can make a weak list look precise. Start with decision labels and short explanations instead:

• Ready: The audience fits, the asset fills a clear need, the page is live, and an appropriate contact route exists.
• Research: One important question remains, such as geographic fit, maintenance status, or replacement suitability.
• Improve asset: The opportunity is relevant, but your resource does not yet meet its needs.
• Exclude: The fit is weak, the placement depends on a link scheme, or the publisher has declined contact.

Within the ready group, prioritize the clearest reader benefit and the least unresolved work. A short submission through a documented resource policy can take precedence over an uncertain pitch to a large publication.

Assess page context and publisher legitimacy using the checks in how to evaluate backlink quality. Domain metrics can help organize a large research pool, but they should not override evidence from the actual page.

Record enough detail that someone else could understand the decision:

Prospect page:
Audience and task:
Specific section or missing resource:
Our proposed destination:
Why readers would benefit:
Evidence checked and date:
Open question:
Contact route or submission policy:
Decision and next action:
Owner:

Avoid collecting personal information you do not need. A public editorial submission form may be enough. Deduplicate at the publisher level as well as the page level so several teammates do not approach the same organization independently.

Know when to stop collecting

Prospecting is unfinished until a shortlist leads to an action. A small team that has enough well-researched candidates for its next outreach cycle should pause collection and test the quality of its reasoning.

Track what happens to the shortlist: how many candidates survive manual review, which exclusions recur, how long verification takes, and what editors say. These observations improve the discovery process. They do not establish how many rankings or conversions a future link will produce.

If most prospects fail on audience fit, narrow the search. If relevant editors consistently explain that your resource duplicates what they already have, revisit the asset. If you cannot identify a reader benefit, adding more prospects is unlikely to solve the underlying problem.

Move the ready candidates into a link building outreach strategy with a specific explanation for each publisher. Save the reviewed table from your Claude or ChatGPT session, with the evidence dates and your decisions. The standard for a finished prospect list is simple: every row should explain who benefits, why this page is relevant, and what happens next.