Guide · Internal links

Do internal links help AI search?

Updated 2026-11-07 · kagentx

Short answer

Yes, at least for crawlers discovering pages: Google's documentation says every page you want found should be linked from at least one other page on the site, using an a tag with href and specific anchor text. Our practice is at least two entry points per new page: one from a list page and one from the body of a related older page. Whether AI crawlers also discover pages through internal links is not stated in any of the three official documents, so it is unverified. The script at the end counts inbound links: on 2026-11-07 each of MochiStamp's 15 pages had 14 inbound pages, while kagentx's old product library has pages with 0 and 1.

What do internal links do for crawlers and AI?

Internal links let crawlers discover new pages and tell them how pages relate. Google's documentation (checked 2026-11-07) is direct: every page you want found should be linked from at least one other page on your site. Whether AI crawlers also discover pages by following internal links, we found no explanation in OpenAI, Anthropic or Perplexity documentation, so it is unverified; but internal links are the most common way for a crawler to reach a new page within a site, so treat them as a basic.

A sitemap answers “which pages exist and when did they change” (see How do you write a trustworthy sitemap lastmod? for how to write it); internal links answer “how do these pages connect and which one matters”. You need both; neither replaces the other.

What kind of link can a crawler follow?

Google generally follows only a elements with an href attribute; other forms “mostly won't be parsed”. Links inserted with JavaScript can be crawled as long as they end up as a plus href, but whether the body itself is server-rendered is covered in Do AI crawlers render JavaScript?.

How it is writtenReliable? (Google documentation)
An a tag with href, absolute or relative pathReliable
An a tag with href inserted by a scriptReliable, as long as the page renders normally
href on a spanNot reliable
An a tag with only onclick and no hrefNot reliable
href written as javascript:…Not reliable

How should anchor text be written?

Anchor text is the visible words of a link. Google's advice: read the anchor text on its own and check it is specific enough; “click here” and “read more” are its examples of weak text, and an empty link is worse; when an image is the link, its alt text serves as the anchor text; keep it short and natural, do not stuff keywords, and do not crowd several links together.

How it is writtenExampleVerdict
Specific: state the question the target page answersHow do you write a trustworthy sitemap lastmod?Good. Readers and crawlers know what is behind the link
VagueClick here, learn moreWeak. Out of context it says nothing
Keyword stuffingAI search GEO SEO optimization ranking toolWeak. Unnatural
Image linkThe img alt reads “MochiStamp size guide”Good. The alt is the anchor text

kagentx guides link to each other using the target page's title (or the first half of it) as anchor text, for example Do changelog pages help AI search? and Canonical tags and duplicate content: how to set them up on a multilingual site so pages are not treated as duplicates.

How do you give every page at least two internal entry points?

This is our content operations rule: when a new page is published, at least two other pages link to it. The first entry point comes from a list page; the second comes from a related older page, written into its body.

  1. 1. List page: make the page appear on a central page automatically

    kagentx's /guides/ list page automatically picks up every file in the guides folder, so a new guide has this one entry point by default. MochiStamp's mochistamp.com/guide/ has the same kind of list.

  2. 2. Older pages: add one sentence in the body of related pages

    Pick the two or three older pages closest in topic and add one sentence where they touch the subject, with the target page's title as anchor text. Links only in the footer or sidebar do not count as entry points for us, because a crawler cannot tell how they relate to the body.

  3. 3. Link back from the new page to older pages

    Put two or three links to older pages in the new page. Links then run both ways, and a crawler can walk from either side.

  4. 4. Multilingual sites: link to the same-language version

    Links in each language version must point to pages in the same language; if a page has no version in that language, do not put a link, and do not link to another language.

  5. 5. Before release, check that every link opens

    Open every link on the new page and on each older page you changed, and confirm there is no 404 and no jump to the wrong language.

A real example: how MochiStamp's guide pages link to each other

On 2026-11-07 we read the source of 15 MochiStamp pages with a script: the guide index /guide/, 7 guides, two use-case pages (animated stickers and personal stickers), 4 tool pages and the ideas page /ideas/. Every page is linked from the other 14: the footer and sidebar carry a list of all guides and tools, so no page is an orphan nobody links to. Note that this counts link occurrences and does not separate links in the body from links in the template.

Mistakes happened too. MochiStamp's deploy log (2026-11-06, v87) has a fix line: the link from the Japanese ideas index to the personal stickers page (/personal-stickers) was corrected, and the Persian animated stickers guide no longer links to a guide that has no Persian version. Both were found and fixed after the 11-06 pages went live, which shows links should be opened one by one before release, especially on a multilingual site.

We have gaps of our own. On 2026-11-07 we read the sitemap of kagentx.com with the same script (335 URLs, counting only links written in the HTML): the fewest inbound links went to /compare (0) and a batch of product pages /p/… (1 each). They belong to the old product library, not to the marketing guides, but it shows this check is worth doing regularly.

How do you check your own internal links?

The script below uses only the Python standard library: it reads every URL in the sitemap, counts how many other sitemap pages link to each one, and prints the 15 with the fewest inbound links. It is what we used to read the two sites above.

import re, sys, urllib.request
from collections import defaultdict
from urllib.parse import urljoin, urldefrag

SITEMAP = sys.argv[1]            # e.g. https://example.com/sitemap.xml
LIMIT = int(sys.argv[2]) if len(sys.argv) > 2 else 500
get = lambda u: urllib.request.urlopen(urllib.request.Request(u, headers={'User-Agent': 'Mozilla/5.0'}), timeout=25).read().decode('utf-8', 'replace')

pages = re.findall(r'<loc>([^<]+)</loc>', get(SITEMAP))[:LIMIT]
known = set(pages)
inbound = defaultdict(set)
for url in pages:
    html = get(url)
    for href in re.findall(r'<a [^>]*href="([^"#]+)', html):
        target = urldefrag(urljoin(url, href))[0]
        if target in known and target != url:
            inbound[target].add(url)

for url in sorted(pages, key=lambda u: len(inbound[u]))[:15]:
    print(len(inbound[url]), url)

Usage: python3 inbound.py https://your-site/sitemap.xml 500. A count of 0 is an orphan page; a count of 1 is one short of two entry points. It counts only a href in the HTML source and runs no scripts, which is exactly what a crawler that does not run scripts sees. For a fuller check see AI visibility audit checklist.

Once the links are in place, the page itself must be worth quoting; see How to write content AI will quote: before and after for how to write it.

FAQ

Are more internal links always better?

Google's documentation says there is no ideal number of links, but too many is a sign to trim. What matters is relevance and specific anchor text.

Does a link only in the footer count as an entry point?

It counts as a link, but it is shared by the whole site and says nothing about which part of the body it relates to. Our practice: besides the list page, every new page also needs a related older page linking to it from the body.

Can anchor text be “click here”?

Not recommended. Google lists “click here” and “read more” as weak examples; anchor text should say specifically what the target page is.

Do AI crawlers discover new pages by following internal links?

We found no explanation in the official OpenAI, Anthropic or Perplexity documentation, so it is unverified. The effect of links on Google is documented; do not treat the AI-crawler case as established.

What should I watch on a multilingual site?

Link to the same-language version; if a page has no version in that language, do not link. On 2026-11-06 MochiStamp fixed one Persian page that linked to a page with no Persian version.

Sources

kagentx

Can crawlers find your new pages?
Let the evaluation tell you

The evaluation checks whether crawlers can get in and whether the text is served directly, then asks ChatGPT, Perplexity and Claude (with web search on) your buyers' questions and sees whether you are mentioned. The Basic plan is $1, and results usually take 2–3 minutes.