Auto-link Suggestions: connect related highlights while you read

The hardest part of building a useful reading system is not making highlights. It is finding the passage you already saved three documents ago and realizing it belongs next to the thing you just highlighted.

Auto-link Suggestions helps with that moment. When you create or open a highlight in Highlyt, the app looks across your existing library for related highlights. If it finds a good match, you see a small suggestion in the highlight popup. Click Link, and Highlyt creates the connection while AI chooses the relationship type.

The result is simple: your highlights stop living as isolated saved passages. They start becoming a connected knowledge graph as you read.

The problem: highlights are easy to save and hard to reconnect

Most annotation tools treat every highlight like a bookmark. You save a sentence, maybe add a note, then move on. That works for one document. It breaks when your library grows across PDFs, EPUBs, Kindle imports, web clips, and YouTube transcripts.

You might highlight a framework in one book, evidence in a research paper, and a counterpoint in an article weeks later. The connection exists in your head, but the software does not know about it. To link those ideas manually, you have to remember the earlier passage, search for the right wording, open the other document, and create the relationship yourself.

Auto-link Suggestions reduces that recall burden. It asks: What have you already highlighted that is semantically close to this new highlight?

How Auto-link Suggestions works

When a highlight is created, Highlyt prepares it for semantic matching in the background. The highlight text and your comment are embedded so the system can compare meaning, not just exact keywords. This keeps the reading flow fast: saving the highlight does not wait on the AI pipeline.

Once the embedding is ready, Highlyt searches for similar highlights from your library. The suggestion list is scoped to what your account can access and excludes the current highlight itself. It also avoids showing connections that already exist, so the panel stays focused on useful next links.

In the popup, each suggestion shows the source document, the source type, the color marker, and a short snippet. A PDF highlight, EPUB passage, web clip, or YouTube-linked highlight can all appear with source context, so you know where the idea came from before you link it.

What the AI classifies

Auto-link Suggestions does not just create a generic edge. When you click Link, the backend compares the current highlight with the suggested highlight and chooses one of five relationship types:

• Supports: the suggested highlight reinforces the current idea.
• Contradicts: the suggested highlight challenges or opposes it.
• Expands: the suggested highlight adds detail or extends the idea.
• Questions: the suggested highlight raises a question or doubt.
• Related: the suggested highlight is meaningfully connected but does not fit the other labels.

If the classifier cannot decide, Highlyt falls back to related. The connection still gets created, and the label stays honest instead of pretending to know more than it does.

Benefits for serious reading workflows

The immediate benefit is speed. You do not have to leave the highlight popup, run a search, or remember the exact phrase from another document. Related highlights appear where the decision happens.

The deeper benefit is compounding context. Every accepted suggestion strengthens your knowledge graph. A passage from a PDF can support a Kindle highlight. A web clip can contradict a book. An EPUB note can expand a research paper. The graph becomes a map of how your reading actually connects, not just a list of things you once marked.

That matters for AI workflows too. When your highlights are connected and labeled, the MCP server can expose cleaner reading context to Claude or ChatGPT. Instead of dumping a flat pile of snippets into an AI chat, you give it source passages plus the relationships between them.

How to use it

1. Create a highlight in a PDF, EPUB, or web clip, or open an existing highlight that is ready for suggestions.
2. Watch for the "related highlights found" toast, or click the highlight and look for the sparkle button in the popup.
3. Open the Related highlights panel to review the suggested passages.
4. Click Link if you want Highlyt to create the connection and choose the relationship type.
5. Click Type if you want to choose the relationship manually instead.

After a link is created, the card confirms the chosen type, such as "Linked as Supports" or "Linked as Expands." The new edge then behaves like the rest of Highlyt's highlight links and can appear in the knowledge graph.

UI and UX notes

The interaction is intentionally quiet. Auto-link Suggestions does not take over the reader, open a blocking modal, or force you to process a list before you keep reading. It uses the same Warm Neutral interface language as the rest of Highlyt: canvas surfaces, ink text, coral accents, rounded cards, and Lucide icons.

The UI follows a progressive disclosure pattern. The sparkle control shows that suggestions exist. The pill tells you how many related highlights were found. The full list appears only when you ask for it. This keeps note-taking, color changes, and deletion from competing with a secondary AI workflow.

Feedback is also visible at every step: loading dots while the system checks, a ready state when suggestions exist, disabled states when there is nothing to show, and a success pill after the link is created. On touch screens, controls stay comfortably tappable instead of relying on tiny hover-only affordances.

What this is not

Auto-link Suggestions is not a system that silently rewires your library. It does not create every possible connection, and it does not replace your judgment. The feature surfaces likely relationships, then lets you decide which ones belong in your graph.

That boundary matters. Your knowledge graph should reflect your thinking. AI can shorten the path to good candidates, but the final link is still yours.

Try it in your next document

The next time you highlight a passage in Highlyt, pause for the suggestion. If the system finds related highlights from your library, open the panel and link the ones that actually belong together.

One link is small. A hundred links become a map. That is the point of connected reading: every highlight you save can make the rest of your library more useful.