CoreAEX Link Finder reads public pages from your sitemap and suggests sentences where an existing phrase could link to another page on the same site. Each suggestion shows the exact sentence, the destination and a one-line reason. You choose the rows you agree with and export them as CSV or JSON. Nothing on your site changes.

Open Link Finder at linkfinder.coreaex.com. It is a public beta and needs no account. It reads public HTTPS pages only, checks your robots.txt first, and sends nothing but GET requests, so it cannot write to your site. The suggestions come from rules, not an AI model, and its privacy page says your pages and drafts are never sent to an AI service.

Use it to find internal linking opportunities in content you have already published, or to plan links for a draft before it goes live. It shows where a link could fit. Whether the link belongs, and whether it helps anyone, is still your decision, and the tool makes no promise about rankings, traffic or AI-search visibility.

Choose Whole Site or One Article

Link Finder has two modes. Pick the one that matches the question you are asking.

  • Whole site suggests links between the pages the scan reads, grouped by source page. Use it to find links that are missing across content you have already published.
  • One article looks at a single page: a live article, or a pasted draft plus the URL it will have. Results come in two tabs: links from the article to your other pages, and links to it from them.

To run a scan:

  1. Choose Whole site or One article.
  2. Enter the website. Link Finder reads https pages only.
  3. Choose how many pages to read: 30, 60 (the default) or 120.
  4. Press Scan. Nothing is read until you do.

A pasted draft stays in your browser. Link Finder sends your browser a list of the site's scanned pages, with their titles, URLs, headings and keywords, and the matching runs on your device. Its API has no field that accepts draft text, so there is nothing to log.

What a Scan Reads, and What It Skips

A scan is a sample of your site, not an audit. Link Finder finds pages through your sitemap, reads the newest entries first, and stops at the page cap you chose or after 90 seconds, whichever comes first.

It also leaves out pages that make poor link sources or destinations:

  • listing pages, such as tags, categories and hubs like an "All resources" index
  • pages marked noindex
  • pages without a readable article body

When the scan finishes, a coverage line reports what it covered: "Read X of Y eligible pages." Read that number before the suggestions. A larger cap reads more pages, finds more suggestions and takes longer. If an older page you care about was not read, scan it directly in One article mode.

Before reading anything, Link Finder checks robots.txt. If the file disallows it, or cannot be read because of a server error or timeout, the scan stops rather than assume permission. Every request identifies itself as CoreAEXLinkFinder/1.0 (+https://linkfinder.coreaex.com/), so site owners can see the visits and block them by name.

Whole-site scans depend on the sitemap; Link Finder does not follow links to discover pages. If your site has no sitemap, use One article mode.

Read a Suggestion Card

Each suggestion is a card that shows where the link would go, which words would carry it and why the tool matched them.

PartWhat it tells you
SourceThe page the link would be added to
Section headingThe section the sentence sits in
SentenceThe exact sentence, with the suggested anchor phrase highlighted
Occurrence noteShown when the phrase appears more than once on the page, so you link the right one
DestinationThe page the anchor would point to
ReasonOne line on why the phrase and the destination were matched

Here is a real card from a whole-site scan of coreaex.com on November 9, 2026. The source was the Query Fan-Out guide, in its worked-example section. The sentence read:

...a technical SEO checklist for SaaS migrations and a free Redirect Map Checker, yet neither appears...

The phrase was not linked on the page. The suggested destination was CoreAEX's article on building the Redirect Map Checker.

The match is sound, but the card is a starting point. CoreAEX also publishes a Redirect Map Checker guide and the tool itself. A reader who has just been told the checker exists may be better served by one of those. Open the destination before you approve a row, and pick the page you want that anchor to point at.

What Link Finder Will Not Suggest

Link Finder is conservative by design. When a candidate could produce a bad or awkward link, it drops it, so you get fewer suggestions with less to clean up. It will not suggest:

  • a link from a page to itself
  • a phrase that is already a link
  • text in navigation, footers or related-links blocks
  • text in headings, quotations or code
  • a phrase cut out of a longer name or compound, such as "Content Gap" inside "Content Gap Scanner"
  • generic anchors such as "read more", or one-word anchors
  • a phrase that fits two destinations equally well
  • more than a few links into the same destination in one scan

The coverage panel shows how many candidates were rejected for each reason. A long list of rejections is not a fault. It shows the rules removing matches you would otherwise have to review and discard.

What Still Needs Your Judgement

The rules match words, not meaning, so two kinds of suggestion need a closer look.

Same words, different subject. A sentence about "content gap" in the sense of text missing from a page's HTML could be matched to an article about marketing content-gap analysis. The phrase is identical; the reader's need is not. Read the sentence, then the destination, and ask whether a reader following that link would land where they expected.

Weak context. A match can sit in a sentence where a link adds little, such as a methodology note. Ask whether a reader at that point would want to leave the page.

One rule covers most borderline cases: if the link only works after you reword the sentence, skip it. A phrase that needs rewriting to carry an anchor was not a natural place for one.

Use Your Export

Select the suggestions you approve, then export them as CSV or JSON. The export is a list of proposed links, not changes; nothing has been edited on your site.

Export before you leave the page. A finished scan is held in server memory for 15 minutes and then discarded. After that, reloading the page will not bring the results back.

FieldWhat it holds
source_urlThe page the link goes on
anchor_textThe words that become the link
passageThe sentence the scan found, so the anchor can be located exactly
target_urlThe page the link points to
idA row identifier for tracking which suggestions were applied or skipped

Before applying anything, check each passage on the live page. If a sentence has changed since the scan, skip that row rather than forcing the link into new wording.

You can apply the links by hand or hand the JSON to an editor or a coding agent. Give clear limits with it. For example:

Apply only the rows in this file. For each row, open source_url and find the passage exactly as written. Link anchor_text inside that passage to target_url. If the passage is not on the page word for word, skip the row and list its id. Do not reword any sentence. Show me every change before anything is published.

Limits of the Beta

Link Finder is a public beta with deliberate limits. They are listed here so you can read a result for what it is.

LimitDetail
Page discoverySitemap only; no link-following crawl
Pages per scan30, 60 (default) or 120, newest sitemap entries first
Time per scanStops after 90 seconds
Scans3 an hour and 10 a day, counted per IP address
What it readsPublic HTTPS pages and sitemaps; no forms, sign-ins or other requests
ResultsHeld for 15 minutes, then discarded

To enforce the scan limits, Link Finder keeps four things per scan for 30 days: a salted one-way hash of your IP address, the hostname scanned, the time and the number of pages read. It keeps no page content, no URLs of pages read and no drafts, and the tool runs no analytics. The privacy page has the full detail, including how to ask for your site to be blocked.

What a scan cannot tell you matters as much. It does not read every page on a large site, it does not check whether a suggested link is the best one available, and it cannot promise any search or AI-search result.

Questions About Link Finder

Does Link Finder change my website?

No. It only sends GET requests, which read pages. Adding a link is something you or your editor do after reviewing the export.

Does it read my whole site?

No. It reads up to the page cap you choose, newest sitemap entries first, and the coverage line says how many eligible pages it read.

Is my draft uploaded?

No. The matching runs in your browser against a list of your scanned pages. The draft text is never sent to CoreAEX.

Does it use AI to choose links?

No. Suggestions come from rules, and your pages are not sent to an AI service.

Try it on one site. Run a 60-page whole-site scan, open the first source page, and approve only the rows that read naturally without rewording.

Open Link Finder

To compare a draft against a page that already ranks, read the Content Gap Scanner guide. To find related subtopics worth linking to, read the Query Fan-Out guide. Or see all free CoreAEX tools.

About the author

Zarko Zivkovic is the founder of CoreAEX, building technical SEO, AEO, and AI-visibility systems for B2B SaaS companies. Connect on LinkedIn.