Mike Litman
Write for the Bookmark
X Corpus Research Content Strategy 2026
WRITE
FOR THE
BOOKMARK
The metric nobody tracks · What it reveals about your writing
The like

I saw it.

Consumption signal. You paused. You approved. You moved on. The algorithm counted it. You forgot it before the next scroll.

Feed consumption
Momentary approval
Algorithmic fuel
Zero intent to return
The bookmark

I need this.

Intent signal. You stopped. You thought: I will come back to this. I will use it. I will share it with someone. That intention does not fade the way a like does.

Reference creation
Persistent intent
Saved to use or share
Compounds over time
Write for the Bookmark · 2026
"
A like is a reaction.
A bookmark is an intention.
Only one of those compounds over time.
The distinction that changes how you write
The Optimisation Trap

Most content is built to be seen.
The question worth asking: is it built to be kept?

Reach. Engagement rate. Follower count. None of these tell you if your writing is becoming an asset.

The Diagnostic · B/L Ratio
"
Bookmarks divided by likes.
Above 1.0: article content. Below 1.0: feed content.
When saves equal likes, the article is working as a reference, not just a reaction.
That parity is the threshold.

X shows bookmark count on every article. Open the article. Check the engagement breakdown. Divide by likes. That number is your diagnostic.
B/L Ratio · Bookmark-to-Like · The Diagnostic
The Research

103 articles. 42 authors.
Every B/L measured.

Not opinion. A corpus. Intentionally biased toward quality: we wanted to study what good looks like, not the average.

106
Articles captured
44
Authors profiled
37
B/L ratios measured
3.68
Highest B/L in corpus
B/L Spectrum

37 articles sorted low to high.
The gap is not random.

B/L 0.0 1.0 threshold B/L 3.68
Feed content (below 1.0)
Article content (above 1.0)
High reach, no saves

172K views.
B/L 0.00.

Product-news content. The algorithm amplified it. Nobody saves a feature announcement for later. Events-reactive content does not age into reference.

Views without intent
News-reactive, not timeless
Zero saves per like
Disappears in 24 hours
Reach plus saves

247K views.
B/L 1.73.

Smaller reach. But 1.73 people saved this for every 1 who liked it. That is 73% more bookmarks than likes. A reference, not a reaction.

Views with intent
Evergreen payload
Cited by later writers
Compounds in search
The Five Patterns · What Works
"
It is not tone.
Not niche. Not follower count.
It is whether the article does something
that earns a return visit.
What separates B/L 3.0 from B/L 0.3 · Five patterns follow

Five patterns. Every Hassid article above B/L 2.0 in this corpus has at least three.

X Corpus · 42 Authors · 37 B/L Measurements · 2026
01.
Embed the asset
The article is the lead magnet. Paste-ready prompt, template, code block. Readers save to use, not to admire.
02.
Design the quotable line
One sentence that redefines a category, not summarises a feeling. Given its own paragraph. The citation-primitive ingredient.
03.
Name where it falls short
The trust-gate. Honest failure list before the recommendation. Readers save articles that do not oversell.
04.
Argument before pitch
The founder-analyst mode. Framework first. Product mention once, at the close only. The argument is the value.
05.
Close as gift
Gift close beats three-tier funnel for save-rate. The two biggest outlier articles in the corpus both close without a funnel.
Pattern 01

Embed the asset.

What it is
Something usable inside the body
A paste-ready system prompt. A framework with named steps. A code block the reader can run today. A journal template with five lines.
The article IS the lead magnet
Why it saves
Readers bookmark to return and use it
A like says "good." A bookmark says "I will need this." The asset creates the intent. No asset, no intent. No intent, no save.
Intent, not approval
Pattern 02 · The Quotable Line
"
"Don't prompt Claude. Let it prompt you."

Category-redefining, not feeling-summarising.
Given its own paragraph.
That is the citation-primitive ingredient.
Hassid · cowork-vs-chatgpt · 6.2M views · The line that earns the save
03
Pattern 03 · The Trust-Gate

Name where it falls short.

Before you recommend anything, list the places your idea fails. Hassid calls it "where it falls short (I promised honesty)." Five honest failure modes before the recommendation. Readers save articles that do not oversell because they trust them enough to return.

Present in Hassid top-5 B/L Confession opener variant: benroy 3.6M Trust earns the save, not the pitch
Pattern 04 · Founder-Analyst Mode
"
Framework first.
Product mention once, at the close only.
The argument is the value.
The pitch is not why people save it.
Founder-analyst mode · Most consistently high B/L pattern in the corpus · liampluglab 3.04, ashwingop 1.1 to 1.25 across all articles
The funnel close

Subscribe. Buy. Follow.

Three-tier paid funnel at the end of every article. Optimises for conversion. Sends the signal: this was a sales vehicle. Readers save sales vehicles less.

Converts readers to buyers
Lower share-rate
Signals the pitch behind the content
Caps the save-rate
The gift close

No ask. Just the idea.

The close lands the argument and stops. No funnel. No CTA stack. The article feels complete, not transactional. The two outlier reach articles in the corpus both close this way.

Higher save-rate
Higher share-rate
Signals confidence in the content
cowork-vs-chatgpt (6.2M), benroy (3.6M)
Write for the Bookmark · The Diagnostic
"
Your B/L ratio does not measure
your performance.
It reveals what you have been
optimising for without knowing it.
B/L as a mirror · The diagnostic is also the prescription
Write for the Bookmark · Personal Verdict
"
Most people track reach.
Nobody tracks return.
The B/L ratio is the only metric that tells you
which one you're building.
MIKE LITMAN · Cultural Capital Labs · 2026
Write for the Bookmark · 2026

Write for
the one
who saves.

Continue from slide ?