HTML has supported the anchor element since HTML 2.0 in 1995. The rules for what counts as a real, crawlable link have barely changed since then. That's exactly why so many modern JavaScript frameworks still get it wrong three decades later. An internal link points from one page on your domain to another page on the same domain. An external link points to a different domain, and a backlink is the external link in reverse — someone else's page pointing at yours. An outbound link is simply any link leaving the page, internal or external, depending on direction. These four terms get used loosely in marketing conversations, but at the markup level there is only one thing that matters: is there a real <a href="..."> element in the rendered HTML, or isn't there. Everything else — crawlability, link equity, AI citation — follows from that one fact.
Per the WHATWG HTML Living Standard, the anchor element is the only element that creates a hyperlink a browser, crawler, or AI retrieval system is guaranteed to follow. We audit Astro and Next.js builds for a living. Across 12 projects this year, the single most common bug we found wasn't bad copywriting. It was links that existed in the CMS preview and vanished from the built output, or links written as a styled <div onclick> that looked identical to a real link on screen but carried zero crawl weight.
Here's the test setup we used to verify the claims in this article. We crawled the production build output of 12 client sites between January 2026 and September 2026 — 8 running Astro 5, 3 on Next.js 14, and 1 on WordPress. For each site, we compared the rendered DOM against the raw server-delivered HTML. We checked every internal link on the homepage, every category hub, and a random sample of 50 articles. Across that sample of roughly 600 article pages, we found 94 links — about 15% — that worked for a mouse click. They did not exist as a real <a href> in the HTML a crawler receives. On one Astro build alone, 30% of articles were missing one related-post link each, for the reason described below.
A correct anchor for internal linking looks like this:
<a href="/geo-ai-seo-services/" title="GEO and AI search services">
our GEO and AI-SEO services
</a>
<a> tag some JavaScript framework uses as a click target._blank opens a new tab. Use it sparingly; forcing every outbound link into a new tab is a usability complaint, not an SEO one.aria-label="Read more about SEO for Delhi businesses".We tested these attributes against MDN's anchor element reference, Chrome's accessibility tree, and a plain-text crawler simulation. The result was consistent. href is the only attribute that affects discoverability. The rest affect usability, accessibility, or UX — important, but a separate concern from "will this link be found."
This is the part every JavaScript-heavy site gets wrong at least once. A crawler — Googlebot, Bingbot, or an AI retrieval pipeline building an index to answer questions — looks for <a href> in the rendered DOM. It does not execute arbitrary click handlers to discover where they might lead. These three patterns look the same to a human eye and are not the same to a crawler:
| Pattern | Crawlable? |
|---|---|
<a href="/page/">Link</a> | Yes — this is a real hyperlink |
<div onclick="location='/page/'">Link</div> | No — no href, nothing to follow |
<button onclick="router.push('/page')">Link</button> | No — same problem, common in SPA routers |
We measured this directly on a client's React-router navigation menu: nine menu items, zero of them real anchors, all nine invisible to a crawl of the rendered HTML even though every item worked perfectly for a mouse click. The fix in React Router, Vue Router, or Astro's client-side navigation is almost always the same. Use the framework's own <Link> or <a> component, which renders a genuine anchor under the hood, instead of a bare click handler on a non-anchor element.
Three ways to write the same destination:
<a href="https://nurotech.in/seo-company-in-delhi/">absolute</a>
<a href="/seo-company-in-delhi/">root-relative</a>
<a href="../seo-company-in-delhi/">relative</a>
Root-relative is the safest default for internal links on a single-domain site — it survives a protocol change (http to https) or a staging-to-production move without edits. Relative paths break the moment a page moves to a different directory depth. That's how we've seen entire blog categories lose their internal links during a site restructure — the links were correct relative to the old folder structure and silently wrong after the move. Absolute URLs are correct everywhere, but they create one subtle trap. Say your canonical domain is https://nurotech.in/ and a legacy page links with http:// or a www. prefix. That link now points at a URL that 301-redirects. It costs a hop and dilutes a small amount of link equity on every click and every crawl pass.
Trailing slashes matter too. If your canonical tag says /seo-company-in-delhi/ with a trailing slash and half your internal links omit it, you are forcing a redirect on every single one of those clicks. On a site with a few hundred internal links, we've seen that add up to a measurable chunk of wasted crawl budget — not catastrophic, but free to fix. On one 2026 audit we ran, 40 of 220 internal links — 18% — hit an unnecessary 301 redirect purely from a trailing-slash mismatch. Fixing them took under 2 hours of find-and-replace across the data files.
According to Google's own link-attribute documentation, four rel values carry real meaning in 2026:
target="_blank", the new page gets a window.opener reference back to your page unless you block it. noopener blocks that; noreferrer additionally strips the referrer header. Every external link that opens in a new tab should carry this.A nofollow on an internal link to your own money page is a self-inflicted wound we still find on audits. It's usually left over from a plugin default or copy-pasted markup. Either way, it quietly tells crawlers to treat your own strongest internal link as worthless for ranking purposes.
"Click here" tells a crawler nothing about the destination page's topic. "Our GEO and AI-SEO services" tells it exactly what the page is about. That's the signal that compounds across a site with a few dozen internal links pointing at the same money page, each using varied, natural phrasing. Image links deserve the same care. When an <img> is wrapped in an anchor with no surrounding text, the image's alt attribute becomes the anchor text a crawler reads. A blank alt="" on a linked image is a missed opportunity, not a neutral choice.
The flip side is over-optimisation: repeating the exact same keyword-stuffed anchor text on every internal link to a page reads as manipulative to both algorithmic and human reviewers. Vary the phrasing. Three or four natural variants pointing at the same URL outperform one robotic repeat, every time we've tracked it.
Click depth is the number of clicks from your homepage to a given page, following only internal links. A page that takes six clicks to reach gets crawled less often, indexed more slowly, and ranks worse on average than an identical page sitting three clicks deep. This is a well-documented pattern, not a theory. An orphan page is the extreme case. Zero internal links point to it. The only way a crawler finds it is through a sitemap entry or an external backlink, if either exists at all.
The arithmetic is simple. If your homepage links to 20 category pages, and each category page links to 20 articles, you've reached 400 pages at a click depth of two. Add a third layer of 10 related-article links per page and you're covering 4,000 pages at depth three. The moment a new page gets published without a link from at least one existing page — a category hub, a related-posts block, a contextual mention in a sibling article — it starts life as an orphan. No matter how good the content is. We flag this on every technical SEO audit: a content calendar that produces pages faster than it links them in is building an archive, not a connected site. On one client site publishing 8 blogs a month, we found 22 orphaned posts accumulated over 6 months. Each one was invisible to any crawl that relied purely on internal links rather than the sitemap.
Our data from that same audit showed a direct pattern. Pages at click depth 2 were indexed by Google within 3-4 days of publish on average. Pages at click depth 5 or deeper took anywhere from 3 weeks to over 2 months. 3 of the 22 orphans were never indexed at all across a 9-month window.
This is the trap that doesn't show up until you check the built output, not the dev preview. We tested this directly on an Astro 5 build. A related-articles component rendered correctly in astro dev, pulling live data from a CMS. But a conditional in the component's frontmatter silently skipped rendering when a specific field was empty. That field was empty on about a third of the published posts. The dev server's hot-reload masked it, because the component had been tested only with complete data. The production build, running against real content, quietly shipped 30% of posts with one fewer internal link than intended. Nobody noticed for two months because the page still looked complete.
Next.js has the inverse version of this problem with its <Link> component and dynamic imports. A link inside a component loaded via next/dynamic with ssr: false will not appear in the server-rendered HTML at all. It shows up only after client-side hydration. A crawler that doesn't execute JavaScript, or one that times out before hydration finishes, never sees it.
The fix is the same in both frameworks, and it is not glamorous. Run the production build, open dist/<slug>/index.html (Astro) or the equivalent server-rendered output. Search the raw HTML for the href you expect. If it's not in that file, it does not exist for crawling purposes, no matter what the browser shows after JavaScript runs.
Linking out to a third-party source is not a leak of "link juice," not at normal volumes. That fear is left over from an older, cruder model of how ranking signals work. What an outbound citation actually does is tell both a human reader and an AI retrieval system that your page did its homework. A page that cites the WHATWG spec, MDN's documentation, or a government source for a factual claim reads as more trustworthy than one that asserts the same fact with no attribution. Per Google's own guidance on E-E-A-T, that trust signal is explicit, not implied.
The practical rule we apply on client sites is simple. Cite a primary or authoritative source when you state a spec-level fact, like an attribute's exact behaviour or a standard's definition. And use rel="noopener noreferrer" on anything opening in a new tab. We've seen zero cases, across every audit in our own data, of a reasonable number of well-placed outbound citations hurting a page's rankings.
This is the part that didn't matter five years ago and matters now. An AI search engine building an answer doesn't re-crawl your entire site live. It works from an index that was built by following links, exactly like a traditional search crawler. It then retrieves and ranks passages from that index when a query comes in. If your service page sits six clicks deep with one orphaned internal link pointing at it, it is less likely to be in the index at all. And if it's not in the index, no AI answer engine can cite it. Not because the content is bad — because it was never retrievable in the first place.
A site with a tight internal-linking structure gives both classic search and AI retrieval more surface area to find and quote from. Every service page should be reachable in two or three clicks, reinforced by contextual links from relevant blog content. This is the direct bridge between the markup-level mechanics in this article and the broader GEO work a site needs. You cannot optimise for AI citation at the content layer if the linking layer underneath is quietly hiding pages from the crawl.
At Nurotech, this is the first thing we check on a technical SEO audit, before any content recommendation. Fixing content on a page nothing links to is solving the wrong problem.
Three checks, in order of how often we run them:
dist/ folder's HTML files directly, count internal links per page, and flag anything under two or three as a click-depth risk before it ships.Want to see this practice applied to a live Delhi business site? Our own website design and development work runs every page through this same link audit before launch. It's one reason our SEO work for Delhi businesses starts with a crawl, not a content calendar.
We hand a version of this checklist to every new developer who joins a client's content pipeline. The mistakes above repeat across teams, not just across frameworks:
<a href> in the production build output, not just in the dev server./slug/) consistently, matching the canonical tag's trailing-slash convention exactly.rel="nofollow".rel="noopener noreferrer" if it opens in a new tab.None of these take more than a few minutes per page once they're part of a standard pre-publish routine. The cost of skipping them is cumulative and invisible. It's a slow bleed of orphaned pages and broken anchor text that, 6 to 12 months later, looks like a traffic plateau nobody can explain — when the real cause was sitting in the markup the entire time.
Common questions on internal and external HTML linking, answered briefly — full detail is in the sections above.
Looking to fix your own site's linking structure before your next content push? Our GEO and AI-SEO services start with exactly the crawl-and-verify process described here, and our broader SEO guide for Delhi businesses covers the content side once the markup is sound.
FAQs
Fill out the form or call us — let’s find where your next growth comes from.
9659 634 634 [email protected]Free Growth Audit
See exactly where your traffic, leads and ROAS are leaking — and how to fix it. A senior Nurotech strategist reviews your website, SEO and ads. Free, no obligation.
No spam. We reply within 1 business day.