What is the difference between inbound and outbound mode?

Inbound finds which existing pages should link TO a target you give it, which is how you build authority to a new or important page. Outbound finds which pages a source you are editing should link OUT to. The agent reads your request to pick the direction, and if it is genuinely ambiguous it asks once before running.

Does it edit my website for me?

By default it returns a reviewable plan: a ranked table with the page, suggested anchor text, the match type, and the exact sentence to update, so you stay in control and never ship links you did not approve. It can also make the edits for you if you connect your CMS through an MCP server. Connect WordPress, Webflow, or any CMS that exposes an MCP server, and the agent can apply the approved links directly.

What tools does the internal linking agent use?

Firecrawl maps your site and reads pages, Parallel extracts the link-worthy passages from each candidate, and web search is built in. Semrush is optional enrichment for ranking keywords and page value. Firecrawl, Parallel, and Semrush are connected automatically when you click **Get started**.

No. Semrush only adds enrichment: it confirms a page's ranking keywords to use as anchors and filters out pages with no organic value. If Semrush is not connected, the agent skips that step and still scores relevance and finds anchors from the page content directly.

How does it avoid suggesting weak or forced links?

It scores every candidate on how naturally a link would fit and drops anything below the bar. It suggests one link per target per source by default, and if nothing meets the threshold it returns an empty table and tells you honestly rather than padding the list.

How many pages does it check at once?

Up to 30 candidates per run by default, so quality stays high and tool usage stays reasonable. If your pool is larger, the agent samples the most relevant pages by topic or asks you to narrow it. You can also hand it an exact list to check only those.

We raised a $50M Series B led by Benchmark

Find every internal link opportunity on your site

Name: AI internal linking agent
Author: Gumloop

Use the AI internal linking agent to find your best inbound and outbound links, with suggested anchor text and the exact passages to update.

Works with

Explore 100+ integrations

Trusted by leading teams

Case study

Stop leaving internal links on the table

The highest-ROI SEO work that nobody has time for, run from a single URL in one chat.

Without Gumloop

With Gumloop

Without vs. With Gumloop

Internal linking is the highest-leverage SEO work, and the first thing everyone skips because it means re-reading the whole site.

Hand the agent one URL. It reads your pages and returns a ranked list of the best links to add, in one chat.

You ship a new landing page and it sits there with zero internal links pointing to it, so it never gets the authority to rank.

Run it in inbound mode. It finds every existing post that should link to the new page and where the link goes.

You are editing a post and know it should link out somewhere, but hunting for the right target page kills your momentum.

Run it in outbound mode on the post. It surfaces the product and solution pages it should link to, with the anchor text.

Most linking tools spit out a giant list of weak matches you then have to sanity-check by hand.

The agent scores every candidate and drops anything weak, so what you get back is only links worth adding.

What is the AI internal linking agent?

The Gumloop AI internal linking agent finds the highest-value internal links on your site and hands you a ranked, reviewable plan to add them. Give it a URL and it works in one of two directions. In inbound mode it finds which existing pages should link TO that target. In outbound mode it finds which pages a source you are editing should link OUT to. Either way you get suggested anchor text and the exact passage to update.

It builds the candidate pool for you. Point it at a domain and it discovers the right pages from your sitemap, or hand it a specific list of URLs and it uses only those. It reads each page with Firecrawl, pulls the link-worthy passages with Parallel, scores every candidate for relevance, and drops anything weak. It will not invent a link that does not make sense, and when nothing is a real fit it tells you honestly instead of padding the list.

The output is a clean markdown table you can act on immediately: rank, relevance score, the page, the anchor text, the match type, and the verbatim sentence where the link belongs. Connect Semrush and it also pulls each target page's ranking keywords to use as anchors and filters out pages with no organic value. The agent is fully customizable in chat: change direction, narrow the candidate pool, or raise the bar on what counts as a strong match.

What you can do with the AI internal linking agent

Workflows the agent handles out of the box.

Inbound and outbound in one agent

Find existing pages that should link to a target, or find the pages a post you are editing should link out to. It reads your intent from the request and picks the direction.

Builds the candidate pool for you

Auto-discovers the right pages from your sitemap by topic, or uses a specific URL list you provide. It respects your list and never silently expands it.

Real relevance scoring, no weak links

Reads each candidate, scores how naturally a link would fit, and drops anything below the bar. When nothing fits, it returns an empty table and says so.

Anchor text and the exact passage

For every opportunity it suggests an anchor that reads naturally and quotes the verbatim sentence to update, classified as ready-to-wrap, light-rewrite, or manual review.

Optional Semrush enrichment

Pulls a page's top ranking keywords to use as anchors and skips targets with zero organic traffic and zero ranking keywords, so you link to pages worth linking to.

How to use the AI internal linking agent

Get from landing page to live agent in a few clicks.

1
Click "Get started"
A preconfigured agent is created in your Gumloop workspace with Firecrawl, Parallel, and Semrush attached, plus web search built in. Takes a few seconds.
2
Connect your tools
Authorize the integrations you want. Firecrawl and Parallel do the reading and extraction. Semrush is optional enrichment, and the agent works with whatever is connected.
3
Give it a URL
Drop in a page URL. The agent confirms the direction, asks whether to use your sitemap or a list you provide, then returns a ranked table of link opportunities.

AI internal linking agent use cases

Real workflows teams run with this agent.

Give a new page the links it needs to rank

Just published a landing page or product page? Run inbound mode to find every existing post that should point to it, so it earns internal authority fast.

Add the right links while you write

Editing a blog post? Run outbound mode on it. The agent surfaces the product, solution, and related pages it should link to, with the anchor and the sentence.

Audit internal links across a content hub

Point it at a cluster of articles and let it map the links missing between them, so your topic cluster actually links together the way Google rewards.

Work from a specific list of URLs

Hand it the exact pages you care about and it checks only those, no sitemap crawl. Perfect for a focused refresh or a single high-value target.

Why use Gumloop for the AI internal linking agent

Preconfigured for internal linking

No agent setup. Firecrawl and Parallel are built in, Semrush is ready for enrichment, and web search is on by default.

Skeptical by design

It scores every candidate and refuses to force weak links. An honest empty result beats a list of links that hurt more than they help.

Quotes real passages, never paraphrases

Every suggestion includes the verbatim sentence from the actual page, so you can find the spot and make the edit without second-guessing it.

Customize as you chat

Flip direction, narrow the candidate pool, raise the relevance bar, or cap suggestions per page, all in the same conversation.

Related agents

SEO keyword research agent

Find the keywords worth ranking for first, ranked by volume and difficulty, before you build the links.

SEO content brief generator

Turn a keyword into a writer-ready brief with SERP research, keyword guidance, and an outline.

AI SEO Optimization Agent

Before you link a page up, get a scored brief of the terms and sections it needs to rank.

With enterprise-grade infrastructure and security

Role-based access control

Manage reusable roles, shared credentials, and secrets with scoped access controls.

Virtual private cloud deployments

Deploy Gumloop in your own cloud.

Anthropic

OpenAI

Gemini

DeepSeek

AI model restrictions

Control which AI models teams can use. Set guardrails and enforce spend policies.

Usage monitoring

Track organization-wide credit usage in real time. Implement budget and quota controls to avoid surprises.

TODAY

KatherineCreated an Agent

AronSet up a Slack Trigger

GonzaloConnected Salesforce

WEDNESDAY

MarceloShared a Private Chat

MaxCreated a New Team

Audit logging

Capture detailed audit trails for actions across the organization to understand where data is flowing.

AI proxy support

Bring your own API keys and route requests through your own proxy.

Single Sign-On

Securely streamline identity and access management.

Zero Data Retention

Gumloop never uses customer data to train AI models. For third-party models, we have Zero Data Retention (ZDR) agreements and Data Processing Addendums (DPAs).

SOC 2 Type II Certified

Gumloop is committed to security, and is compliant with SOC 2 Type II and GDPR. Visit trust.gumloop.com to learn more.

Frequently asked questions

Tasks automated to date

9876543210.,

Try the AI internal linking agent

Talk to Sales