Science of Highlighting: Why Most People Do It Wrong

Does Highlighting Actually Work?

Yes, but only when you do it well. Highlighting the way most people do it, dragging a marker across half the page, produces almost no learning benefit. Highlighting done selectively, with a clear system and a follow-up review, is a genuinely effective study technique. The tool was never the problem; the technique is.

In 2013, psychologist John Dunlosky and four colleagues published a landmark review in Psychological Science in the Public Interest. They evaluated ten common study techniques and judged how well each one's benefits hold up across different learning conditions, students, materials, and test types. Highlighting and underlining landed in the lowest tier: "low utility."

The media loved it. "Stop Highlighting Your Textbooks!" headlines spread. Education blogs repeated the verdict as gospel. Teachers told students to put away their highlighters. Within a few years, "highlighting doesn't work" became conventional wisdom.

But most people never read the actual paper. They read a summary of a summary, and the nuance got lost along the way.

The Dunlosky review evaluated highlighting as it is typically practiced, not as it could be practiced. And how do most students highlight? They drag a yellow marker across nearly every line on the page, creating a false sense of familiarity without any real engagement with the material. That's the version of highlighting that doesn't work.

The question nobody bothered to ask was: what happens when you highlight well?

---

What Dunlosky Actually Said

Let's look at the fine print. Dunlosky et al. (2013) concluded that highlighting and underlining, as most students use them, "do not consistently boost students' performance," and can even hurt performance on higher-level inference questions by fragmenting the text.

That word "consistently" matters. It means there were cases where highlighting did boost performance. The review wasn't a blanket condemnation; it was a warning that most students use the technique poorly.

The authors acknowledged a critical limitation: most of the studies they reviewed simply asked participants to read and mark text without providing any training on how or what to mark. Participants received no guidance on selectivity, no instruction on combining highlights with other strategies, and no framework for reviewing highlighted material.

Imagine evaluating the effectiveness of cooking by observing people who've never been taught to cook. You'd conclude that cooking doesn't produce good meals. But the problem isn't cooking; it's the lack of skill.

Dunlosky and his colleagues rated five techniques as "low utility" (highlighting, rereading, summarization, keyword mnemonics, and imagery use) and only two as "high utility" (practice testing and distributed practice). The high-utility techniques shared something important: they forced active processing. The low-utility techniques, as typically used, were passive.

This distinction between passive and active is where the real insight lies. Highlighting can be passive, yes. But it doesn't have to be.

---

When Highlighting Works: The Research

Since 2013, a growing body of research has complicated the "highlighting is useless" narrative. Several studies have identified conditions under which highlighting becomes genuinely effective.

The clearest summary comes from a meta-analysis. Ponce, Mayer, and Méndez (2022), writing in Educational Psychology Review, pooled 85 effect sizes from 36 studies. Learner-generated highlighting improved memory (d ≈ 0.36), though not comprehension (d ≈ 0.20), and it helped college students (d ≈ 0.39) more than younger school students (d ≈ 0.24). When an instructor provided the highlights, both memory and comprehension improved (d ≈ 0.44 each). Read that again: expert-chosen highlights help even more than self-made ones, which tells you the skill of choosing what to mark is doing the heavy lifting.

A telling result comes from Yue, Storm, Kornell, and Bjork (2015), also in Educational Psychology Review. They found that students who didn't believe highlighting was useful actually benefited more from it than the enthusiasts, and that highlighting didn't impair memory for the non-highlighted material. The benefit came from engaging with the text, not from believing in the ritual.

More recent work extends this to screens. Mason and Ronconi and colleagues (2024) found that in digital reading, the quality of a student's highlights predicted both literal comprehension and transfer, even when the raw amount of highlighting didn't.

What ties these positive findings together? In every case, highlighting worked because it was combined with a deliberate process: selectivity, annotation, or structured review. The highlighter wasn't the study tool. The thinking behind the highlighting was the study tool.

---

The Selectivity Principle

The single biggest predictor of whether highlighting helps or hurts your learning is how much you highlight.

Fowler and Barker (1974), in a Journal of Applied Psychology study titled "Effectiveness of highlighting for retention of text material," ran one of the earliest tests of this and found no reliable overall benefit from highlighting as students typically did it. Later work sharpened the picture: mark too much and the passages you highlight lose their advantage over the ones you skip. If you highlight everything, you highlight nothing. The visual contrast that makes highlighting useful, separating important from unimportant, disappears when 70% of the page is yellow.

Contrast this with selective highlighting. Reading specialists recommend highlighting only after you've finished reading a paragraph or section. The process should be: read first, think about what matters, then go back and mark only the core idea. One sentence per paragraph is a good starting point. Two at most.

This aligns with how expert readers naturally engage with text. Skilled academics don't highlight as they read; they highlight during a second pass, after they've understood the structure of the argument. The highlight becomes a bookmark for retrieval, not a substitute for comprehension.

Here's a simple test for whether you're highlighting well: if you come back to a page a week later, can your highlights alone reconstruct the argument? If so, you've highlighted selectively. If your highlights just form a wall of color with no clear thread, you've fallen into the passivity trap.

Tools like Glasp's web highlighter support this kind of selective engagement. When you highlight a passage on any webpage, you're making a conscious decision about what matters. And because Glasp stores your highlights in a personal library, you can revisit them later for review, turning a one-time reading act into a spaced retrieval practice. Learn more about how to highlight text on web pages effectively.

---

Color-Coding: More Than Aesthetics

If selectivity is the first principle of effective highlighting, a color system is a useful second. Using different colors to categorize information adds a layer of processing that single-color highlighting misses.

The evidence on color and memory is softer than study blogs suggest, so it's worth being careful about what's actually established. Color can aid memory in some conditions: a review by Dzulkifli and Mustafa (2013) in the Malaysian Journal of Medical Sciences concluded that color can improve memory performance, partly by boosting attention and arousal. But claims that a specific color scheme raises recall by a precise percentage usually trace back to uncited blog posts, so treat exact figures with skepticism. The defensible point is simpler: assigning colors to categories forces you to classify information as you read, and that classification is a form of elaborative processing.

A common system codes four colors by meaning:

• Yellow: Main argument or thesis
• Blue: Methods and evidence
• Pink/Red: Points you disagree with or want to question
• Green: Connections to other readings or your own work

Academic reading specialist Raul Pacheco-Vega uses a different logic, coding by level of importance rather than content type: yellow for the key ideas, then pink, green, and blue for progressively lower-tier points. Either approach works. What matters is that you assign meaning to each color before you read and stay consistent. Over time, your brain starts categorizing information automatically as you read, and the classification itself becomes a form of elaborative processing, which asks more of you than dragging a marker across a line.

Glasp supports multiple highlight colors, making it easy to implement a color-coding system across all your web reading. You can assign meaning to each color and maintain consistency across articles, papers, and blog posts.

---

Digital vs. Paper: What Changes?

For years the assumption was that paper always beats screens for serious reading. Highlighting complicates that picture.

In a 2024 study of digital reading, Lucia Mason and colleagues found that what mattered wasn't whether students highlighted but the quality of what they highlighted. Students whose highlights captured the important ideas scored higher on both literal comprehension and transfer (Mason & Ronconi et al., 2024, Journal of Computer Assisted Learning). Precise selection, not the medium, drove the benefit. And precise selection is exactly what digital tools make easier.

Digital highlighting offers several advantages over paper:

Precision. Digital highlighting lets you select exact phrases and sentences, while paper highlighters often bleed into surrounding text. There's no accidental over-marking.

Searchability. Paper highlights are useful only when you physically return to the book. Digital highlights can be searched, tagged, organized, and exported. With Glasp, all your highlights are automatically saved to your profile and can be searched across your entire reading history.

Portability. Your highlighted passages travel with you. If you highlighted an article on your laptop, you can review those highlights on your phone during a commute. Glasp even supports Kindle highlights, bringing your physical reading into your digital knowledge library.

Integration with AI. This is a capability that paper simply can't match. Glasp's AI Summary feature can analyze your highlights and generate summaries, identify themes, or suggest connections between passages you've marked across different articles. This turns your highlights into inputs for deeper thinking, not just passive bookmarks. You can read more about how AI is reshaping the way we learn.

That said, digital reading comes with its own challenges. Screen fatigue is real. The temptation to skim is stronger on screens. And some research finds no significant difference between digital and paper comprehension overall. The point isn't that digital is universally better; it's that digital highlighting tools, used intentionally, remove several friction points that make paper highlighting less effective.

---

Highlighting vs. AI Summaries: Do You Still Need to Read?

A new question has appeared that the older highlighting studies never had to face: if an AI can summarize an article in seconds, why bother marking it up yourself? It's a fair question, and the honest answer is that AI summaries and active highlighting solve different problems. One saves you time. The other builds understanding. They're strongest together and weakest as substitutes.

Start with what decades of memory research established long before ChatGPT. The generation effect, first documented by Slamecka and Graf (1978), shows that information you produce yourself is remembered better than the identical information you simply read. Deciding "this sentence is the point" is a small generative act. Accepting a machine's summary is the passive "read" condition. Retrieval practice points the same way: Karpicke and Blunt (2011), writing in Science, found that students who actively recalled a text learned more than students who studied it with elaborate concept maps, even though the students predicted the opposite. Reading an AI summary is closer to passive re-exposure than to active recall. And Robert Bjork's concept of desirable difficulties explains why the effortless option so often loses: conditions that make learning feel harder in the moment tend to produce more durable memory. An AI summary optimizes for exactly the fluency that under-builds retention.

The newest evidence is suggestive but should be read with caution. A 2025 MIT Media Lab preprint (Kosmyna et al., "Your Brain on ChatGPT") recorded weaker, less distributed brain connectivity in people who used an AI to write essays than in those who wrote unaided, and found the AI group often couldn't quote what they'd just produced. It's a small study (54 participants), not peer-reviewed, and about writing rather than reading, so treat it as an early signal, not proof. Two other 2025 findings converge: a Microsoft and Carnegie Mellon survey of 319 knowledge workers (Lee et al., CHI 2025) found that the more people trusted an AI tool, the less critical thinking they reported doing, and a study of 666 people (Gerlich, 2025) found frequent AI use correlated with lower critical-thinking scores, mediated by cognitive offloading. Both are self-reported and correlational, so they show an association, not that AI makes anyone think worse.

None of this makes AI the enemy of learning. It means the reader's judgment is the part worth protecting. The workflow that respects the science looks like this: you read and highlight selectively, doing the cognitive work of deciding what matters, and then you hand those highlights to AI to summarize, cluster, and connect. Glasp is built for exactly this loop. You highlight what's important, and Glasp's AI works from what you chose rather than from the raw article, so the synthesis stays grounded in your own judgment. If you want a deeper look at the study-focused tools, see our guide to AI study modes compared and our breakdown of NotebookLM in 2026. Let the human do the selecting. Let the machine do the synthesizing.

---

Social Highlighting: The Multiplier Effect

Here's where things get interesting. Everything we've discussed so far treats highlighting as a solo activity: you read, you mark, you review. But what if you could see what other people highlighted in the same text?

This is the idea behind social highlighting (sometimes called collaborative annotation), and the research behind it is encouraging.

Cornell University's Center for Teaching Innovation describes social annotation as "reading and thinking together," bringing "the age-old process of marking up texts to the digital learning space while making it a collaborative exercise." Quasi-experimental studies of tools like Perusall back the practice up: students who complete social-annotation assignments tend to score higher on exams (Miller et al., 2018; Suhre et al., 2019). These results are correlational rather than randomized, but they point consistently in the same direction, because students don't just passively absorb information; they actively reflect on their reading in conversation with a community.

The benefits break down into several categories:

Distributed attention. No single reader catches everything important in a text. When you see highlights from dozens or hundreds of other readers, you discover passages you might have skimmed over. It's like having a study group that reads everything for you and flags the best parts.

Social proof of importance. If 200 people highlighted the same sentence, that's a strong signal that the sentence captures something essential. This form of crowd-sourced curation helps readers prioritize, especially in unfamiliar subjects where they might not know what to look for.

Exposure to different perspectives. Other readers' annotations reveal how people from different backgrounds interpret the same text. A data scientist might highlight the methodology section of an article that a designer would skip entirely. Seeing both perspectives enriches your own understanding.

Accountability and motivation. Knowing that others can see your highlights creates a subtle motivation to read more carefully and highlight more thoughtfully. It's the same principle behind why people exercise more consistently when they have a workout partner.

This is the core of what Glasp offers. Glasp's community feed lets you see what other readers have highlighted across the web. You can follow people whose reading tastes align with yours, discover new articles through their highlights, and build on each other's knowledge. It's highlighting as a social learning practice, not just a personal study hack.

Glasp also organizes highlights by topic, making it easy to explore what the community has found most valuable in any subject area. Whether you're researching machine learning, philosophy, or product design, you can browse topic-based collections of highlights from readers around the world.

---

A Practical Highlighting Protocol

Based on the research, here's a step-by-step protocol for highlighting that actually improves learning:

Step 1: Read First, Highlight Second

Never highlight on your first pass through a paragraph. Read the entire section, understand the argument, and then go back to mark the key point. This prevents the most common mistake: highlighting before you know what's important.

Step 2: Limit Yourself

Aim for one highlight per paragraph, two at most. If everything feels important, that's a sign you need to re-read with a clearer question in mind. Ask yourself: "If I could only remember one thing from this section, what would it be?"

Step 3: Use Color with Purpose

Assign meaning to each color before you start reading. A simple system:

Stick with this system consistently. Over time, your brain will automatically start categorizing information as you read.

Step 4: Add a Note

For every two or three highlights, write a brief annotation. This can be a one-sentence summary, a question, or a connection to something else you know. The annotation is what transforms passive highlighting into active processing. Glasp lets you attach notes to any highlight, keeping your thinking right next to the source material. For a complete guide to annotation across all media types, see our guide to annotation.

Step 5: Review and Retrieve

Highlights are useless if you never look at them again. Schedule a weekly review of your recent highlights. Tools like Glasp make this easy by collecting all your highlights in one place, organized by article, date, or topic. During review, try to recall the context of each highlight before re-reading the source. This acts as a form of retrieval practice, one of the highest-rated learning techniques from Dunlosky's own research. Learn more about how active recall works.

Step 6: Share and Discuss

Share your highlighted passages with others. Post them to your Glasp profile, discuss them with colleagues, or use them as starting points for writing. Teaching and explaining what you've read is one of the most effective ways to solidify understanding.

---

Passive vs. Active vs. Social Highlighting

Here's how the three approaches compare across key dimensions:

The progression from passive to active to social represents a shift from marking text to engaging with ideas to learning in community. Each level builds on the previous one.

---

Frequently Asked Questions

Is highlighting really a bad study strategy?

No, but passive highlighting is. The 2013 Dunlosky review that popularized this claim specifically evaluated highlighting without selectivity, annotation, or structured review. When students highlight selectively (one to two sentences per paragraph), use color-coding, and add marginal notes, highlighting becomes a genuinely effective learning technique. The tool isn't the problem; it's how most people use it.

How much should I highlight on a page?

Less than you think. Fowler and Barker (1974) found no reliable overall benefit from highlighting as most students do it, and the visual contrast that makes highlighting work breaks down when most of the page is colored. A practical rule of thumb is to highlight no more than 10 to 20% of a text. For a typical paragraph, that means one sentence, possibly two. If you find yourself highlighting more than a third of the text, step back and re-read the section with a specific question in mind.

Does the color of my highlighter matter?

A little, but not in the way you might expect. Individual colors have modest, and somewhat contested, effects on attention. The real benefit comes from using a system of colors, where each color represents a category of information. That forces you to classify what you're reading as you read it, which is a form of elaborative processing that strengthens memory encoding.

Should I still highlight if AI can summarize the article for me?

Yes, and the two work best together rather than as substitutes. Deciding what matters is the cognitively active step that builds memory (the generation effect; Slamecka & Graf, 1978), and handing that judgment to an AI means skipping it. The stronger workflow is to read and highlight actively yourself, then use AI to summarize your highlights, surface what you missed, and connect ideas across sources. Let the human do the selecting and the machine do the synthesizing.

Is digital highlighting better than paper highlighting?

Digital highlighting has real advantages: precise selection, searchability, portability, and integration with AI. In a 2024 study of digital reading, the quality of what students highlighted predicted their comprehension and transfer (Mason & Ronconi, 2024). But the medium matters less than the method: selective highlighting combined with annotation and review is what drives learning, on screen or on paper.

How does social highlighting improve learning?

Social highlighting works through several mechanisms. First, it exposes you to passages you might have overlooked, expanding your attention. Second, seeing that many other readers highlighted the same passage serves as a signal of importance. Third, reading others' annotations introduces you to different interpretations and perspectives. Quasi-experimental studies of collaborative annotation tools like Perusall consistently link the practice to better exam performance, though these results are correlational rather than randomized.

Can I use Glasp for academic research?

Absolutely. Glasp lets you highlight any web page, attach notes, and organize your highlights by topic. For academic research, you can use color-coding to categorize findings (evidence, methods, counterarguments), export your highlights for use in papers, and use the AI Summary feature to identify patterns across multiple sources. The community feed also helps you discover relevant articles through the highlights of other researchers in your field.

---

Conclusion: Highlighting, Done Right

The "highlighting doesn't work" myth is one of the most persistent oversimplifications in learning science. What the research actually shows is more nuanced and far more useful: mindless highlighting doesn't work, but strategic highlighting is a powerful learning tool.

The evidence is clear. Selective highlighting forces you to evaluate what's important. A consistent color system adds a layer of categorization that strengthens memory. Adding annotations transforms a passive mark into an active thought. And sharing your highlights with a community creates a social reinforcement loop that benefits everyone involved. Even in the age of instant AI summaries, the one step worth keeping for yourself is the judgment about what matters.

These aren't complicated techniques. You don't need to overhaul your entire study routine. You just need to shift from "mark everything that looks important" to "read first, think second, highlight third, annotate fourth." For strategies on reading more deeply, see our guide to deep reading.

Glasp was built for exactly this kind of intentional reading. It's a free web highlighter that lets you highlight any page on the internet, organize your highlights with colors and notes, review them in a personal library, and share them with a community of curious readers. Whether you're a student, researcher, or lifelong learner, Glasp turns your reading into a lasting knowledge asset.

Stop highlighting everything. Start highlighting what matters. And let others see what you found.

Get started with Glasp for free →