Plan Link Building Outreach with Claude or ChatGPT

A well-written email cannot rescue an irrelevant suggestion. Before you work on a subject line, establish why a particular editor's readers would benefit from the page you want to share.

That explanation is the core of a link building outreach strategy. Research identifies a real editorial opening, your resource fills it, and the message gives the publisher enough information to decide. Sometimes the right result is a link. Sometimes it is useful feedback, a correction, or a clear refusal that helps you stop pursuing a poor fit.

Use Claude or ChatGPT with RankParse to research the context and draft a specific proposal, then have a person verify and send it. The process below includes a fictional campaign and adaptable prompts and messages, not a promise of placements or ranking gains.

Give the campaign a reader-facing purpose

“Get more backlinks” is a goal for your team. It does not explain the benefit to a publisher.

Describe the campaign in terms of the problem your resource solves. For example:

Help organizations supporting independent repair businesses give their members a practical way to estimate workshop capacity.

That statement tells you what to build, whom to research, and what an appropriate placement might look like. It also excludes broad business sites that have no relevant audience.

Define the destination page and the limits of its promise. A capacity worksheet should explain its inputs and assumptions. If it cannot account for specialist staffing, emergency repairs, or regional working-time rules, say so. An editor should not have to discover these limitations after recommending it.

Before outreach, check that the page loads, works on mobile, identifies its author or responsible organization, and delivers what the pitch will describe. For a downloadable asset, test the actual download. For a calculator, check representative inputs and edge cases. A resource that is hard to use creates avoidable work for the publisher and reader.

Build a brief for each prospective publisher

Start with a small group of candidates you can research properly. There is no useful universal target for daily sends: the limiting factor is whether each suggestion has a credible editorial reason.

For every candidate, capture:

• The exact article or resource page you are referencing.
• The audience it serves and the task readers are trying to complete.
• The section where your resource might help.
• What existing references already provide.
• The additional value your resource offers.
• Any submission rules, contact preferences, or restrictions.

Do not confuse a topic match with an opportunity. An article on running a repair shop may already have a stronger capacity template, or may focus entirely on customer acquisition. In either case, your worksheet may not belong there.

Use how to find link building opportunities to turn discovery into a shortlist. Then evaluate the page and publisher before investing time in a tailored message. These checks should remove weak candidates, not generate flattering sentences for a generic pitch.

Build the campaign packet in Claude or ChatGPT

Connect through the secure OAuth route in the RankParse MCP guide, then enable the connector for your research conversation. The web connector is read-only. It can supply backlink evidence and live page metadata, but cannot discover contacts through outreach tools, send messages, or create approval-inbox drafts. The workflow's output is a reviewed document for your normal email process.

Use one packet per asset and audience. Here is a fictional example; replace the domains and inputs before running the prompts:

Company: RepairLedger, scheduling software for repair workshops
Sender: [real name, role, and relationship to the company]
Asset: https://repairledger.example/resources/capacity-worksheet
Audience: independent repair workshop owners
Benefit: compare bookings with realistic available technician hours
Proof: editable worksheet; worked example; assumptions explained
Access: no account required, download tested by [reviewer/date]
Limit: does not model employment law or specialist staffing needs
Competitor domain: shopplanning.example
Shortlisted publisher: repairassociation.example
Candidate page: https://repairassociation.example/start-a-workshop
Contact policy: unknown until manually checked
Campaign history: [publisher owner, last contact, reply, opt-out]
Output: research brief, draft if justified, reviewer checklist

Supply the resource itself or the relevant page text when the assistant cannot browse it. Calling a worksheet “free” or “editable” requires evidence from the actual asset; a backlink row cannot prove either claim. Include previous correspondence so that the assistant does not write a first-contact message to someone who already declined.

Stage 1: retrieve the citation that explains the prospect

A shortlist often says only that a publisher links to a competitor. Use RankParse to recover the page-level context before writing. Query get_backlinks for the competitor, narrowed to the publisher's exact from_domain:

Using the campaign packet, 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. Identify any
row matching our candidate page. If none matches, say so; do not
replace it silently with another page or call it a verified opening.
Explain which competitor resource was cited and what still requires
reading the page. Return the evidence brief only, not an email.

Use sort=recent because the default importance ordering may return aggregated rows whose from_url is a homepage. An aggregated row proves neither where the reference appeared nor what the editor said. The backlinks documentation describes both shapes. Recent means newest within the available snapshot, not newly published today.

If the publisher has many sampled citations, group them by exact referring page and destination. A staffing article citing a hiring checklist is a different opportunity from a startup guide citing a business-plan template. Do not blend their passages into an email about a nonexistent page.

Stage 2: check the publisher page and your promise

Use get_page_seo with the exact public URL of the proposed placement and then the destination asset page. It gives you data.status_code, data.final_url, data.title.text, data.headings, and other page metadata. Check that the final page is the one you mean to reference and that the destination has not become an error page or a gated landing page.

Call get_page_seo for [publisher page URL] and [asset page URL].
Record status_code, final_url, title.text and relevant headings.
Identify redirects or error responses that need review. Do not
infer that a particular citation exists from links.external.

Prepare questions for my browser check: the exact current passage,
existing equivalent resources, submission policy, current contact
route, and whether our asset fulfills every proposed claim.

This is a page check, not the final editorial review. links.external is a count, not a list of destinations. A successful tool invocation can contain a target status_code of 404. The fetch reads raw HTML, so client-rendered content or a bot challenge may require browser inspection. Follow the page SEO reference if a URL fails; fragments must be removed from tool inputs, although you can retain the section name in your campaign notes.

Read the publisher's page and submission guidance yourself, or use a separate browsing capability and verify the retrieved passages. Record the source URL and check date. Neither an existing competitor citation nor a visible contact address means the publisher invites submissions. The checked policy determines the route, while the article determines the reader benefit.

Stage 3: transform checked evidence into an editorial brief

Here is an invented evidence trail for RepairLedger. These values illustrate how to reason from a response; they are not a live tool result:

The useful brief is: “Offer the worksheet in the scheduling section because it connects bookings to available hours.” It is not “you linked to our competitor, so link to us.” The competitor citation helped you find and understand the page; the proposed contribution needs its own justification.

Give the verified notes back to the assistant:

Here are checked passages, submission instructions, campaign history
and asset checks, each with source and date: [paste].

First decide draft / hold / exclude. Hold if the reader benefit or
contact route is unknown. Exclude if our history contains an opt-out
or the resource duplicates an adequate existing reference.

If drafting is justified, write one concise message for the stated
submission route. Identify my role and the exact section where the
resource helps. Map each factual claim in a separate review table to
my notes or tool evidence. Do not invent praise, results, personal
interests, email addresses or prior conversations. Save the output
in this chat; do not send or create a proposal.

For the fictional evidence above, the assistant should produce a form submission draft and a claim table. A row for “no account required” points to the asset check; “your scheduling section” points to the current passage. The crawl date stays in the research record, not in the message as a claim about when the editor added the link.

Stage 4: make the handoff ready for a reviewer

Before transferring a draft into your email or form, challenge it:

Audit this draft against the evidence brief. List every statement
that lacks a source, overstates the resource, or implies familiarity
we have not established. Remove those statements and show a revised
draft. Preserve any unresolved question as a hold reason.

Return a campaign row with publisher, exact page, proposed resource,
reader benefit, evidence dates, verified submission route, owner,
review status, prior-contact status and next action.

For RepairLedger, an illustrative row might be Repair Association; startup guide / scheduling; capacity worksheet; helps compare bookings with technician hours; form verified by reviewer; owner: Sam; draft ready for human review; no prior contact recorded; next action: review form submission. Keep the actual draft beside that row. “Draft ready” is not “sent,” and a tool-created research brief is not an approved message.

A solo operator can manage this in one sheet and work through a few publishers at a time. A larger team needs a shared publisher-level owner and contact history before anyone sends. Two asset teams may have different ideas for the same association; coordinate those ideas into one editorial conversation.

Troubleshoot the brief before rewriting the email

If you later choose the API-key route, RankParse documents a separate outreach integration requiring its own setup. Its existence does not make those actions available in the Claude or ChatGPT OAuth workflow described here. Keep this guide's deliverable as the reviewed research packet and message draft.

Choose the contact route before the wording

Follow the publisher's stated submission process. A resource submission form, corrections address, or named section editor is often more suitable than a founder's personal inbox.

Check whether the person still holds the role and whether that role covers the page. A historical byline does not necessarily identify the current owner. If no suitable route is available, record the gap and move on rather than contacting several unrelated employees.

Within your team, assign one owner per publisher. Keep a record of contact dates, responses, and requests not to be contacted. Respect applicable email and privacy requirements, and stop when someone asks you to stop.

This coordination is especially useful when several writers or agencies work on the same site. An editor who receives three versions of the same suggestion has been given three tasks, not three reasons to consider it.

A fictional campaign, from asset to message

Imagine a fictional company, RepairLedger, creates scheduling software for repair workshops. It has published a free capacity worksheet that owners can use without a software subscription. The worksheet distinguishes booked work from realistic available technician hours and explains how to include an allowance for unplanned jobs.

A researcher finds a trade association's guide to planning a first workshop. Its staffing section explains hiring, and its finance section provides a startup budget. It does not offer a way to compare planned bookings with available working hours. The association accepts suggestions for member resources through an editorial address.

The proposed benefit is narrow: help readers connect staffing and booking decisions. The researcher confirms that the worksheet uses understandable language, does not require an account, and does not imply that its output replaces professional financial advice.

Here is an example message for that fictional situation:

Subject: Capacity worksheet for your workshop planning guide

Hi [editor's name],

Your workshop planning guide covers staffing and startup costs.
In the section on scheduling, readers may also need a way to compare
planned bookings with the technician hours they can actually offer.

I work at RepairLedger. We published a free capacity worksheet that
makes that comparison and includes an editable allowance for unplanned
jobs: [resource URL]. It works without an account, and the assumptions
are explained on the page.

Would it be useful as an optional resource in that section? If a
different format would be more helpful for members, I'd welcome
that feedback.

[Name and role]

The message identifies the existing page, the missing task, the sender's relationship to the resource, and the proposed use. It makes a small, specific request. It does not demand an exact anchor text or imply that inclusion is owed.

If the association already has a suitable worksheet, the pitch should not be sent unchanged. The team needs a distinct contribution or a different prospect.

Adapt the request to the editorial situation

Use a template to remember what evidence to include, not to replace research. Different opportunities need different requests.

For a broken reference, say when you checked it and avoid presenting a temporary loading error as a permanent failure. For an attribution request, confirm that the publisher is actually referring to your work.

If you are sharing research, include its publication date, method, and relevant limitations. Do not describe a small convenience sample as an industry-wide finding. The more readily an editor can verify the contribution, the easier it is to assess.

A concise email can link to the full method rather than reproducing it. Concision should remove unnecessary words, not evidence needed to understand the claim.

Plan follow-ups as a service, not a sequence

A non-response is ambiguous. The person may be busy, may not maintain the page, or may have decided the suggestion is unsuitable. It is not permission for an indefinite series of reminders.

For a small campaign, one follow-up is a reasonable starting policy unless the publisher's preferences say otherwise. Allow a meaningful interval that fits their publishing pace; a weekly newsletter and an annual resource review operate differently. Do not follow up after an explicit refusal or opt-out.

A useful follow-up keeps the original context visible:

Hi [editor's name],

Following up once on the capacity worksheet for your workshop
planning guide: [resource URL]. The relevant use is helping readers
compare bookings with available technician hours.

If it is not a fit for your resource list, no action is needed.

[Name]

Add a substantive update only when one exists. Do not invent urgency, pretend you have a prior relationship, or imply that you can see the recipient avoiding your message.

If there is still no reply, close the prospect. Reopen it only when there is a meaningful new reason, such as a publisher inviting submissions or your resource changing in a way that addresses a documented concern. A new month is not itself a new reason.

Make outcomes useful to the next decision

Count the results that help improve the work. Sent-message totals mostly describe activity. Open tracking is an uncertain signal because privacy features and automated systems affect it; it is a poor basis for judging interest.

A simple campaign record can separate these outcomes:

• Not a fit: The publisher explains an audience, scope, or format mismatch.
• Needs improvement: The resource has a useful premise but lacks something specific.
• Accepted or published: The publisher chooses to reference the resource.
• Declined or opted out: Further contact stops.
• No response: The prospect is closed after the planned follow-up policy.

Also record research time, repeated objections, and whether a published reference points to the right destination. If a link sends referral traffic or relevant enquiries, report those observations with their date range and attribution limits. Avoid presenting subsequent ranking movement as proof that the placement caused it.

SEO reporting that drives decisions can help put these observations alongside other work and business outcomes. A useful campaign conclusion might be “association editors want a printable version,” which leads directly to an asset improvement. “We sent more emails” offers much less guidance.

Set the boundary before negotiations begin

Some publishers will respond with a price list, a request for a reciprocal link, or an offer to guarantee placement across several sites. Decide in advance how your team handles these responses.

A genuine advertising arrangement should be evaluated as advertising: audience fit, disclosure, cost, and appropriately qualified links. Google explains the relevant attributes in its guidance on qualifying outbound links. Do not purchase a followed placement on the assumption that it guarantees ranking value.

An editorial campaign should also leave the editor free to choose wording, placement, qualification, or no inclusion at all. If the arrangement depends on controlling those details for ranking purposes, revisit the purpose of the campaign.

Keep the verified brief and final message beside the RankParse evidence in your campaign record. The finished strategy is a set of defensible choices: a useful asset, a researched audience, a clear request, a stopping rule, and a way to learn. Start with the publishers whose readers have the clearest need. Their decisions and feedback will tell you more about the quality of the campaign than a larger contact list ever could.