How to Turn Your X/Twitter Liked Posts into an AI Knowledge Base

Most advice about rescuing your X/Twitter content from the platform focuses on bookmarks. But bookmarks aren't where most people actually save things. A lot of X usage looks like this: you're scrolling, a thread teaches you something, you tap the heart, and you move on — meaning to come back to it later. Months later, your Likes tab is a scroll of hundreds of tweets with zero structure, no search that actually works, and no way to ask "what did that thread about pricing strategy actually say?"

X doesn't make this easier. There's no "export my likes" button anywhere in the app, and the few third-party tools that exist are built to dump your likes into a raw CSV or JSON file — useful if you want a spreadsheet, useless if you want to actually ask questions of that content. PostToSource exists for exactly this gap: turning the social content X locks away into clean documents you can feed to NotebookLM, ChatGPT, or Claude.

Why Liked Tweets Are Harder to Deal With Than Bookmarks

Bookmarks and likes solve different jobs, which is part of why they end up in different states of disorganization:

• Bookmarks are usually intentional — you tapped "save for later" because you meant to act on it. Likes are often just a reflex. That means a Likes tab mixes genuinely useful research material with jokes, replies to friends, and posts you liked just to signal support.
• Likes are public by default on your profile (unless you've hidden the tab in settings), while bookmarks are always private — which shapes what people feel comfortable liking versus bookmarking, but doesn't make either one easier to search.
• Neither has a native export, but likes get even less tooling attention than bookmarks because the use case is seen as "vanity metric," not "personal archive."
• Liked posts can vanish the same way bookmarked ones do — if the original poster deletes the tweet, locks their account, or gets suspended, your like points at nothing. The content you meant to revisit is just gone.

If you've already sorted through why bookmarks disappear, the same instability applies to likes — arguably worse, since you're less likely to have the tweet text saved anywhere else.

Step 1: Get Your Liked Posts Out of X

X's own interface only lets you scroll your Likes tab — there's no bulk export in the standard app. Here's what actually works, roughly in order of effort:

Option A: Manual collection (best for under ~50 posts)

1. Open your profile and go to the Likes tab.
2. Scroll through and copy the URLs of posts worth keeping. X loads likes in reverse-chronological order, so recent ones are fastest to grab.
3. Drop the URLs into a spreadsheet or a plain text file as you go.

This is tedious past a page or two, but it's free, requires no third-party access to your account, and works even if your likes are set to private.

Option B: Browser extension or export tool (for a full backlog)

A handful of extensions and tools (listed in X's own ecosystem of third-party apps) can scrape your profile's Likes tab and output a CSV or JSON of liked tweet URLs and text. If you go this route:

• Only use tools that run locally in your browser rather than ones that ask for your X password or full account access.
• Expect raw output — a spreadsheet of URLs, timestamps, and sometimes tweet text, not a readable document.
• Treat the "and media" and "and followers" extras most of these tools advertise as noise you don't need for a knowledge base.

Option C: X's own data archive (full history, slow)

X's account settings include a Download an archive of your data option under Settings → Your Account. It includes your like history as part of the full account export, but it arrives as a large structured data dump (not a tidy document) and can take a day or more to generate. It's the most complete option, but overkill if you just want last quarter's liked threads turned into research material — see how to work with a full X data archive if you're going this route anyway.

Step 2: Filter Before You Build Anything

Unlike bookmarks, your Likes tab is not a curated list — it's a mix of everything you've ever tapped a heart on. Before you feed it to an AI tool, cut it down:

• Drop anything that isn't informational. Jokes, replies to friends, and support-signaling likes add noise without adding knowledge.
• Group by topic, not by date. A knowledge base organized as "AI tooling," "marketing frameworks," "industry news" is far more useful to query than one organized as "everything I liked in March."
• Re-check that the content still resolves. Deleted tweets and locked accounts mean dead links — confirm what's still live before you spend time structuring it.

Step 3: Turn the Surviving Links into AI-Ready Content

Once you have a filtered list of URLs, raw links still aren't something NotebookLM, ChatGPT, or Claude can use well on their own. PostToSource handles the conversion:

1. Paste your list of liked-post URLs into PostToSource, one at a time or in a batch.
2. PostToSource fetches each post's text, author, timestamp, and any linked media, and formats it into a clean content block.
3. Export the result as Markdown or plain text, ready to drop into an AI tool as a source.

This turns "a spreadsheet of tweet links I liked eight months ago" into structured documents an AI tool can actually reason over.

Step 4: Import Into NotebookLM (or Your AI Tool of Choice)

1. Open NotebookLM and either create a new notebook or add to an existing research notebook.
2. Click Add source and upload your cleaned Markdown files, or paste the content directly.
3. If you're importing a large backlog, check NotebookLM's source limit first so you don't hit a wall halfway through.
4. Ask NotebookLM to summarize themes across your liked posts, pull every mention of a specific topic, or cross-reference them against other sources in the same notebook.

The same cleaned Markdown works just as well pasted into a ChatGPT or Claude project if that's your tool of choice — the formatting work is what makes it portable.

Practical Tips

• Do this in batches, not all at once. A Likes tab with two years of history is a bigger filtering job than it looks. Pull a few months at a time and build the knowledge base incrementally.
• Combine with bookmarks. Likes and bookmarks capture different kinds of content people meant to keep — pairing both, alongside your saved X threads, gives a fuller picture than either alone.
• Unlike as you archive. Once a liked post is safely converted into your knowledge base, consider unliking it on X — it keeps your public Likes tab from becoming an unreadable scroll of everything you've ever found mildly interesting.
• Revisit your X list follows too. If you're systematically mining X for research, Twitter/X Lists are worth the same treatment.

Frequently Asked Questions

Can I export all my liked tweets from X in one click?

No. X has no native bulk export for liked posts. You can collect links manually, use a third-party browser extension that scrapes your Likes tab, or request a full account data archive from Settings, which includes likes but arrives as a large structured file rather than a readable document.

Are liked posts private on X?

By default, your Likes tab is public and visible on your profile, though you can hide it in your privacy settings. Either way, visibility doesn't affect export — there's still no built-in way to pull the list out as a document.

What happens to a liked post if the original tweet gets deleted?

It disappears from your Likes tab, or stays listed but links to nothing. The same instability that affects bookmarks applies to likes — if you want to keep the content, you need to capture it before the original poster deletes it, locks their account, or gets suspended.

Is it worth building a knowledge base from likes instead of just bookmarks?

If you use bookmarks intentionally and likes more casually, bookmarks alone will cover most of your deliberate research saves. But if you're like most X users and tap "like" far more often than "bookmark," your Likes tab actually holds more of your informational history — it just needs more filtering before it's useful.

Conclusion

Liked posts on X hold real research value, but the platform treats them as a vanity feature rather than an archive — there's no export button, no structure, and no guarantee the content stays live. Pull your likes out manually or via a scraping tool, filter out the noise, and convert what's left into clean, structured documents. PostToSource automates that cleanup step, so turning months of liked posts into a queryable AI knowledge base takes an afternoon instead of becoming a project you keep putting off.