How to remember vocabulary you read on the web: a capture-and-review system that scales

You read English (or French, German, Japanese...) on the web every day — news articles, research papers, blog posts, documentation. You see unfamiliar words. You think "I should remember that." You don't remember it.

A week later, you encounter the same word again. You look it up again. Same forgetting cycle. After a year, you've looked up "ubiquitous" or "préalable" or "それゆえに" twenty times — and still feel uncertain when you'd actually use it.

This article describes a system that breaks the cycle. It works for any language, any reading goal — IELTS prep, academic research, leisure reading, professional content consumption. The core idea: capture in context, review with spaced repetition, measure by output.

Why "I'll remember this one" doesn't work

The brain prioritises encoding for things that survive two filters:

Attention — did you process the item with conscious effort?
Re-encounter — did you see it again before forgetting?

Just thinking "I should remember this" passes filter 1 weakly and almost always fails filter 2. Within 24 hours, 65% of newly-encountered words are gone (Ebbinghaus, 1885 — confirmed by hundreds of replications). After a week, 75%. The forgetting curve is mathematically brutal and indifferent to your good intentions.

The fix isn't willpower. It's a system that captures words at the moment of encounter and surfaces them again on a schedule that beats the curve.

The 3-component workflow

Component 1 — Capture at the moment of encounter

This is non-negotiable. If your workflow involves "I'll add this to my study list later," you've already lost. Later doesn't come.

The minimum viable capture flow:

Read in your normal browser.
Highlight an unfamiliar word.
See definition + IPA + audio + translation in an in-page popup.
One click to save into a flashcard system.

Total time: under 3 seconds. Doesn't break reading flow.

This is exactly what Mnemo's Chrome extension does — and importantly, the extension automatically saves the enclosing sentence along with the word. Context comes free, no manual transcription.

A non-negotiable property: the saved data must auto-sync to a review system you'll actually open. Browser bookmarks fail this — you never look at them. A separate notes app fails this — you never review notes. A dedicated flashcard system that schedules reviews succeeds.

Component 2 — Curate weekly (15 minutes)

Capture is high-volume and low-quality on purpose. You're not trying to make perfect cards in the moment — you're trying to not lose anything. Curation comes later.

Once a week, open your inbox of saved words and:

Add 1-2 collocations — for substantiate: substantiate a claim/argument with evidence. The collocation is what makes the word usable, not the definition.
Write one sentence that's relevant to your life or work. This personal anchor makes the word stick.
Tag by category — Work, Reading, Domain X. Tags help you study by topic later.
Delete junk — words that are too common (you knew them) or too rare (won't appear again). Be ruthless. A 200-word focused deck beats a 2000-word bloated one.

Fifteen minutes once a week. That's it.

Component 3 — Review daily, briefly

Open the app, review what's due. Modern spaced repetition algorithms (FSRS v5 in Mnemo, also available in recent Anki) schedule each card based on your individual performance — you don't decide when to review, the algorithm does.

Daily review properties that matter:

15-20 minutes max. Beyond that, fatigue degrades encoding quality.
Retrieve before flipping. Always try to recall the meaning before checking. Active retrieval is roughly 50% more effective than re-reading (Karpicke & Roediger, 2008).
Rate honestly. "Good" should mean you actually retrieved naturally. Lying to the algorithm causes intervals to stretch too far → real forgetting → frustration.

After 4 weeks of daily review, your due-card count stabilises around 30-60 per day. This is the steady state — sustainable indefinitely.

What about workflow alternatives?

A few common alternatives and why they fall short:

Manual flashcard creation (typing cards by hand). High friction → you don't do it consistently. Even when you do, you typically write just "word = meaning" — losing the context.

Underlining in a physical book or PDF. Same problem as "I'll remember this one" — re-encounter never happens unless you build a separate review habit. Underlining feels productive but doesn't survive the forgetting curve.

Browser bookmarks / Pocket / Notion clips. Great for full articles, terrible for individual words. The granularity is wrong.

Anki + a third-party browser add-on. Works if you're already an Anki power user, but the add-ons are unmaintained and friction-prone. The capture-to-card flow is still 4-5 clicks.

Quizlet manually. Works for short-term learning (one course in one semester) but the SRS algorithm is shallow and the paywall has gotten aggressive.

The dedicated highlight-to-flashcard tool with built-in spaced repetition is the only workflow where all three components are connected without friction.

Sample workflows by reader type

The news reader (general English improvement)

Daily input: 30 min of BBC, Guardian, NYT.
Capture: ~10-15 words/day during reading.
Curate: Sunday morning, 20 minutes.
Review: 10 minutes every weekday morning.
Output check: every 2 weeks, write one 200-word piece using recently-learned words.

The researcher (academic papers in second language)

Daily input: 1-2 papers in your field.
Capture: ~20-30 words/day, heavy on field-specific terminology.
Curate: end of each paper, ~10 minutes (re-read your highlights, add field-specific collocations).
Review: 15-20 minutes morning + 5 minute commute review.
Output check: after 3 months, write a paragraph synthesising recent reading using captured vocabulary.

The polyglot (multiple languages in maintenance mode)

Daily input: native content in 2-3 languages (podcast for one, article for another, etc.).
Capture: 5-10 words/language/day.
Curate: weekly, separate decks per language.
Review: rotation system — heavy review for the "active" language, light maintenance for others.
Output check: monthly conversation/writing in each language.

The exam student (IELTS, TOEFL, GRE, etc.)

See the dedicated articles:

The same workflow applies — exam prep is just one slice of "vocabulary from reading".

Common pitfalls

Capturing too much without curating. A 5,000-card unreviewed inbox is paralysis. Cap weekly intake at ~50 new cards. Quality > quantity.

Skipping review days. Skip one day, the schedule self-corrects. Skip a week, you have a backlog of due cards that triggers "abandon the system" thoughts. Daily ≠ negotiable.

Lying to the algorithm. Tapping "Good" on cards you struggled with feels nice in the moment but breaks the schedule. You'll forget the word for real later, then blame the system instead of the dishonest rating.

Treating it like productive procrastination. Card-creation can become busywork. The point isn't to have cards — it's to have words you can use. Measure output (writing, speaking, comprehension) every 2 weeks; if output isn't improving, your inputs aren't sticking and you need to reduce volume + increase review depth.

Tools

For this workflow specifically, you need three things working together:

A capture tool (Chrome extension or browser bookmarklet)
A spaced repetition engine (FSRS or modern equivalent)
Something connecting the two without friction

Options:

Mnemo — built specifically for this workflow. Chrome extension is a first-class citizen, FSRS v5 by default, free, open source. Multi-language. Try free.
Anki + Yomichan/Anki Web Importer — if you're a power user willing to configure. The community add-ons work but aren't maintained at the same level.
Quizlet — only if your goal is short-term and you accept the paywall + shallow SRS.

For most readers building a long-term vocabulary system, Mnemo is the lowest-friction path because the components were designed together rather than bolted on.

Summary

If you're a reader who wants vocabulary from your reading to actually stick:

Capture at the moment — no "save for later", no underlining without review.
Curate weekly — collocations + personal sentence + delete junk.
Review daily — 15-20 min, honest ratings, retrieve before flipping.
Measure by output — every 2 weeks, are you using the words?

That's the whole system. The hard part is consistency, not strategy.