Recently Updated: 17 min read

Link Evidence in Browser DevTools: href, rel, Canonical

Link Evidence in Browser DevTools: href, rel, Canonical

A link counts as evidenced when three readings agree: the href and rel values on the anchor, the canonical and robots meta tags in the head of the page that carries it, and a dated note saying which view of the page was read. A screenshot of the Elements panel records that reading visually, but a screenshot alone does not hold the date, the address or the view.

If the interface is set to another language, menu names can look different.

What a developer tools reading can and cannot prove

The limits decide how a note should be worded. A DevTools reading is a record of one browser session looking at one page at one moment.

What the reading shows

QuestionAnswered by DevTools?What stays open
Is the link on the page right now?Yes, at the moment of readingWhether it is there next week
Are href and rel what the contract says?Yes, in that sessionWhether another visitor gets the same page
Did Google crawl the page?NoCrawl records
Did Google process the link?NoRendered view as Google sees it
Is the page in the index?NoIndex status

How the note should be worded

Practical suggestion: write the claim as "on this date, in this browser, the page showed this link with these values". Wording such as "Google counts this link" has no support in a DevTools reading. Google's crawlable links documentation (Opens in a new tab) says that, for links inserted with JavaScript, the rendered anchor text can be checked with the URL Inspection tool. That tool belongs to the property owner in Search Console. A link buyer who does not own the publisher's property can only ask the publisher for that check, and the DevTools reading becomes the reason for asking.

The documentation has no statement on how a link seen in DevTools is counted in Google's index. The note therefore stops at what the page showed.

Two views of one page: HTML and DOM

The Chrome DevTools DOM documentation (Opens in a new tab) separates two things that are easy to blur. HTML is what the server returns. The DOM is the tree of objects the browser builds from it.

The HTML the server returned

In the documentation's wording, HTML represents the initial page content. A link that sits in that text is in place before any script runs, so it is a record that depends less on scripts. Practical suggestion: keep a line copied from the browser's page-source view as its own entry in the note.

The DOM once scripts have run

The same page states that when JavaScript adds, removes or edits nodes, the DOM becomes different from the HTML. Its example is a script that removes a heading and adds a paragraph. A link can follow the same path: a script can add it, rewrite its address, or take it out. The Elements panel shows the DOM, the current state of the page after the browser has done its work.

Google's crawlable links documentation adds that links inserted into a page dynamically with JavaScript are crawlable when they use standard anchor markup with an href. A link that appears only in the DOM is therefore not a problem in itself. The open questions are whether its markup is standard and whether the values match what the buyer paid for. The article on whether a purchased link is crawlable goes deeper into link markup.

Which view goes into the file

Practical suggestion: store both. If the two readings match, the link does not depend on a script. If they differ, note exactly which attribute differs and ask the publisher why.

FindingPage sourceElements (DOM)Reading
Same link, same valuesPresentPresentNo script dependence, record complete
Link only in the DOMAbsentPresentAdded by script. Check markup shape and, through the owner, the rendered view
Link only in the sourcePresentAbsentA script may remove it. Ask the publisher
rel differs between viewsOne valueAnother valueThe final value is unclear. Record both

According to the DevTools documentation, the DOM Tree of the Elements panel is where all DOM-related work in DevTools is done. Menu items are named in English: "Elements", "Inspect", "Select an element" and "Edit as HTML".

Right-click and "Inspect"

The documentation describes the fastest route: right-click an item on the page and choose "Inspect". The Elements panel opens with that node highlighted in the DOM Tree. When the expected option is missing from the context menu, the documentation suggests right-clicking away from the node's text.

The "Select an element" tool

This tool lets a reader pick a node on the page and find it in the tree. It suits links that are hard to hit with a right-click, such as a small icon inside a sidebar.

Example scenario: an archery range's feature article on a local sports blog carries the link on the phrase "beginner lessons". Inspecting that phrase highlights the anchor, and the relevant attributes appear in the panel.

Reading the href

Google's crawlable links documentation says Google can't reliably extract URLs from anchor elements that don't have an href attribute. Its recommended form is an anchor with a real address in the href. Not recommended in the documentation are an href placed on a span, a javascript: address, and an anchor with an onclick handler but no href. An onclick handler beside a real href is listed as recommended, so the handler itself is not the issue.

An href is read twice. First for shape: does the value resolve to an actual web address? Then for target: is it the page that was bought, or another one? Differences in parameters, fragments and letter case are covered in the target URL accuracy article. Practical suggestion: copy the value as text and store it apart from the screenshot.

Reading rel values

Google's documentation on qualifying outbound links (Opens in a new tab) defines three values. "sponsored" marks advertisements and paid placements. "ugc" is recommended for user-generated content such as comments and forum posts. "nofollow" is for cases where the other values don't apply. Several values may be given as a space-separated or comma-separated list, and the documentation's examples are "ugc nofollow" and "ugc,nofollow".

The documentation says links marked this way will generally not be followed, while the linked pages may be found through other means. The note records rel as found: the values as written, or "no rel attribute".

Which value belongs on a purchased link depends on the agreement between buyer and publisher. Knowing that Google recommends "sponsored" for paid placements is a starting point for that conversation. A schedule for repeating the rel check is set out in the monitoring calendar article.

The documentation asks for anchor text placed between anchor elements that Google can crawl. An empty anchor, a generic phrase such as "click here" and overly long text are on its list of poor examples. As a fallback, Google can use the title attribute as anchor text if the element is empty. For an image that is a link, Google uses the alt attribute of the img element as anchor text. A linked image with empty alt and no other text therefore has no anchor text at all.

FieldWhere to lookWhat to recordWatch for
hrefAttribute list of the anchorFull address, as textMust resolve to a real web address
relSame anchorThe value or "no rel"Multiple values may be space or comma separated
Anchor textChild node of the anchorVisible textEmpty or generic text is a reason to ask for a fix
Alt text on an image linkalt attribute of the imgThe alt valueAlt works as the anchor text
TargetThe href itselfAddress of the bought pageParameters and fragments are judged separately

Head checks: canonical, robots meta and header rules

A correct link can sit on a page whose head changes the page's own status. The head is therefore part of the evidence.

The canonical element

According to Google's documentation on consolidating duplicate URLs (Opens in a new tab), a rel="canonical" link element is only accepted if it appears in the head section of the HTML. The same page calls it a strong signal that the specified URL should become canonical, and it recommends absolute paths rather than relative ones. In the Elements panel, open the head node and look among the link elements for the one whose rel is canonical.

Three facts go into the note: whether a canonical exists, its exact address, and whether that address equals the address of the page the link sits on. A canonical that points to a different address suggests the page is not marked as the main version. Fixes and audits for that case are covered in the canonical tag issues article. If the head holds two different canonical elements, record both as found and pass them to the publisher.

The robots meta tag

Google's robots meta tag documentation (Opens in a new tab) says to place the tag in the head section of a page. The value "noindex" means do not show this page, media or resource in search results, and "nofollow" means do not follow the links on this page. To address a specific crawler, the robots value of the name attribute is replaced with the crawler's token, and Google supports two such tokens in the tag; other values are ignored.

Header rules the panel cannot show

The same rules can arrive through the X-Robots-Tag element of the HTTP response header. That header is part of the server's response, not of the HTML, and the Elements panel shows only the DOM built from HTML. Seeing no noindex in the panel therefore does not prove the response header carries none. The noindex and X-Robots-Tag guide goes through both forms.

Interplay with robots.txt

The documentation states that URLs containing indexing or serving rules cannot be disallowed from crawling, because otherwise the rules will not be found and will be ignored. A page carrying a meta tag but blocked in robots.txt leaves the tag's effect uncertain. Reading the publisher's robots.txt is outside DevTools, and the robots.txt audit article lists the steps.

Reading the findings: what each one calls for

Decision table: finding, interpretation, next step.

FindingInterpretationNext step (practical suggestion)
Anchor has a real hrefMatches the form Google recommends for crawlable linksStore the value as text
No href, onclick onlyA form the documentation does not recommendAsk the publisher for a real href
href starts with javascript:A form the documentation does not recommendAsk for a standard link
No relNo qualificationCompare with the agreement
rel="sponsored"The paid-placement markerCompare with the agreement
Canonical points elsewhereThe page is not marked as the main versionAsk the publisher why
Robots meta noindexPage is not meant to appear in search resultsAsk whether it is intentional
Robots meta nofollowLinks on the page are not meant to be followedAsk whether it is intentional

A rel on one anchor and a nofollow in the robots meta are separate records. One covers a single link, the other every link on the page. The note lists them on separate lines.

Closing checklist for one log entry

  • The link was searched for in both page source and DOM.
  • The href was copied as text and compared with the target.
  • The rel value, or "no rel", is written down.
  • Anchor text, or the alt value for an image link, is noted.
  • The canonical address from the head is written down.
  • The robots meta value, or its absence, is written down.
  • The note says the X-Robots-Tag header needs its own check.
  • Nothing was edited in the Elements panel.
  • Date, time zone and browser version are recorded.
  • Text copy and screenshot are stored in the same folder.

Applied to one case: a sports-blog feature

Example scenario: the local sports blog tells the archery range that the feature is live. The range owner, or the agency handling its link placements, runs the check in this order.

Step by step

  1. Open the article and write down the address exactly as the address bar shows it.
  2. In the page-source view, search for the range's domain. Copy the anchor's href and rel as text.
  3. Back on the page, right-click the phrase "beginner lessons" and choose "Inspect". Compare the anchor in the Elements panel with the copied line.
  4. Open the head node. Note the canonical element and any robots meta tag, or note that neither exists.
  5. Take one screenshot per view and attach each to the note.
  6. If something differs, for example the source shows the lesson page and the DOM shows a redirect address, write the difference down and send it to the blog editor.

Reading the result

Example scenario: the source shows an href to the beginner-lesson page, rel is "sponsored", the canonical is the article's own address and there is no robots meta tag. All four records agree. That "sponsored" matches the value Google recommends for paid placements; whether it matches the deal is for the buyer to check against the contract.

Different outcome: the canonical points to another article on the blog, or the robots meta carries "noindex". The link itself is right, yet the editor gets two separate questions: why does the canonical point elsewhere, and was the noindex added on purpose?

Keep hands off the attributes

The DevTools documentation states that the DOM can be edited on the fly to see how changes affect the page. Double-clicking an attribute name or value edits it, double-clicking text changes it, a tag name can be retyped, and "Edit as HTML" opens a node's markup for editing. That is handy for debugging. For evidence it creates risk.

Practical suggestion: when collecting evidence, do not edit anything in the Elements panel. An accidental double-click on an href, or a deleted rel attribute, makes the screenshot show an edited state rather than the page. If an edit happened, reload the page and start the reading again.

If an experiment is needed, for instance to see whether a link becomes visible under another rule, keep experiments and evidence in separate sessions. No screenshot from an experiment session goes into the evidence file.

Evidence log layout

A screenshot without context proves little: it does not say when, from which address or in which browser it was taken. The log entry beside it closes those gaps.

Fields for every entry

  1. Full page address as shown in the address bar.
  2. Date, time and time zone of the reading.
  3. Browser name and version.
  4. The href as text.
  5. The rel value, or "no rel".
  6. The canonical address and the robots meta value, or the fact that neither exists.
  7. The view read: page source, DOM, or both.

Text copy and screenshot together

Practical suggestion: add the relevant HTML line to the file as plain text next to the image. A screenshot is a visual record, but it cannot be searched or pasted, and a small attribute is easy to misread in it. A text copy shows a one-letter difference in an href or a typo in a rel value immediately.

A short naming pattern is enough: date, publisher domain, view. For example "2026-11-20-publisher-domain-dom". If the pattern stays the same for every check, a file opened months later explains itself.

Session preparation

Practical suggestions on browser conditions: use the same browser and window type each time. Installed extensions can alter the DOM, so a profile without extensions or a guest window ties the reading to the page rather than to an add-on. A page can look different to a signed-in visitor, and new elements can appear after a cookie banner is dismissed. Record "not signed in" when that is the case, and keep the order of reading fixed: address, source link, DOM link, head, screenshots.

Writing to the publisher

The note's main job is to support a precise request when something is wrong. A message reading "our link is broken" gives the publisher nothing to work with. A short note holding the values read points to the cause.

Fields for the request

FieldExample scenario value
Page addressFull address of the blog feature
Reading date2026-11-20 with browser version
Expected hrefThe range's beginner-lesson page
Observed hrefExpected in the source, a different address in the DOM
Expected relThe value in the agreement
Observed relSource and DOM listed separately
Canonical and robots metaThe value read, or "none"

Tone and follow-up

Practical suggestion: write observations, not accusations. "The DOM shows this href, the source shows that one" gives the publisher's technical contact a direct lead. After the publisher changes the page, repeat the reading in the same order and under the same conditions, and keep the new entry beside the old one.

A repeat reading checks all four fields, not only the corrected one. Fixing one value can change another. If nothing differs, the note closes with "no change".

Mistakes to avoid

Trusting the panel for header rules

No noindex in the DOM does not mean no noindex in the response header. The header needs a separate check.

Shortening the address

An href or page address can carry parameters or a fragment. A trimmed address cannot be compared with the publisher's intended target, so the value is copied in full.

Skipping alt on a linked image

If the range's logo serves as the link, Google's documentation says the alt value is read as anchor text. A linked image with empty alt and no other text has no anchor text. The alt value belongs in the record.

Treating one screenshot as proof for a whole period

A reading captures one moment. A second reading, taken with the same steps after some weeks and stored beside the first, shows whether anything moved.

Reading only the anchor

Canonical and robots meta work at page level and can change the page's status even when the link is right. Head reading is a line on the checklist.

Frequently Asked Questions

On its own, no. The image shows neither the date, the address nor the view that was read. Practical suggestion: store the HTML line as plain text with date and address, and attach the screenshot to that note.

The Chrome DevTools documentation says HTML is the initial content the server returns and the DOM is the current content, and that the two differ when JavaScript adds, removes or edits nodes. When a difference appears, record both values and send them to the publisher.

The DOM of that session has no rel attribute. Search the page source too and record the two results on separate lines. Google's documentation recommends "sponsored" for paid placements, which gives the buyer a reference point if the agreement involves payment.

Which browser version should be recorded?

Practical suggestion: record the browser name and the version shown in its about page, together with the date and time zone, in the same log entry as the href. Note whether the reading ran in a profile without extensions, because extensions can alter the DOM.

Practical suggestion: treat the framed page as a page of its own. Record the address of the frame, read the anchor inside it in source and DOM in the same way, and note in the log that the link sits in a frame. Open the framed address on its own as well and compare.

What if the page needs a login?

Practical suggestion: read the page the way an ordinary visitor meets it, signed out, and write "not signed in" in the note. If the link appears only after signing in, record exactly that and ask the publisher whether the placement is meant to be public.

No. The documentation has no statement connecting a link seen in DevTools to Google's index. A note should say what the page showed on the date of reading and nothing beyond that.

Conclusion

Link evidence rests on three records that support one another: the href and rel of the anchor in both source and DOM, the canonical and robots meta read from the head, and the dated note next to the image. DevTools makes the reading quick, and it stops short of showing what Google crawled.

To plan editorial placements whose delivery you can check this way, explore the backlink packages.

0 Comments

No comments yet for this article.

Post a comment

Please log in to leave a comment. Login