The Don't Make Me Think summary in one line: a visitor should understand a page without working it out. Steve Krug's book argues that people scan instead of read. They click the first link that looks right, and they leave when a page makes them puzzle. Clear labels, obvious buttons and fewer words fix most of it.
Key facts about the book
| Full title | Don't Make Me Think, Revisited: A Common Sense Approach to Web (and Mobile) Usability |
| Author | Steve Krug, usability consultant, Chestnut Hill, MA |
| Edition and publisher | 3rd edition, New Riders, 2014, 212 pages |
| Sales | More than 700,000 copies in 15 languages, per Krug's own book page |
| Companion book | Rocket Surgery Made Easy (New Riders, 2010), on running your own usability tests |
| Who it's for | Anyone who makes a site or app: owners, marketers, writers, designers, developers |
The facts in this table come from the author's book page. What follows is our summary in our own words, plus the checklist we hand a client before a redesign.
What are the main ideas in Don't Make Me Think?
The main idea is the author's First Law of Usability, which is also the title: a page should be self-evident. Everything else in the book explains why that matters and how to get there. Here are the ideas we think carry the most weight, in our words.
People scan, satisfice and muddle through. Visitors don't read a page top to bottom. They glance at it, pick the first option that seems good enough ("satisficing"), and keep going even when they don't understand how the site works. Design for that behaviour, not for the careful reader you wish you had.
Design pages like billboards. Use the conventions people already know (logo top left, navigation where they expect it). Make the most important thing the most prominent. Make it plain what you can click. Break text into chunks someone can scan.
Make choices mindless. His second law is about clicks. The number of clicks matters less than how much thought each one takes. Three easy clicks beat one click that makes you stop and wonder which box you belong in.
Omit needless words. Cut the welcome paragraph that says nothing. Cut instructions that nobody reads. A shorter page makes the useful words easier to find.
Navigation is street signs. Every page should tell you what site you're on, what page this is, what the main sections are and where you are among them. His "trunk test" checks this: imagine you've been dropped on a random inner page, and see if you can answer those questions at a glance.
The homepage has to answer the first questions fast. What is this? What can I do here? Why here and not somewhere else? A clear tagline and a visible starting point do most of that work.
Stop arguing, start testing. Teams waste hours debating what "users like." His answer is to watch a few real people use the site and let that settle it.
Usability is courtesy. Visitors arrive with a limited supply of goodwill. Hidden prices, forced sign-ups and hard-to-find phone numbers drain it. Answering the questions people actually have refills it. The 2014 edition also adds chapters on mobile and accessibility.
Website audit checklist: Krug's rules, one 5-minute test each
You can audit a site against these ideas in under an hour, without software. Each row below takes about five minutes on one page. Run it on your homepage and on your two most-visited service or product pages.
| Principle | What to check on your site | How to test it in 5 minutes |
|---|---|---|
| Self-evident pages | A newcomer can tell what the page is for without reading body copy | Show the page to someone outside the company for 5 seconds, hide it, and ask what it was for |
| Scanning, not reading | Headings, bold lead-ins and short paragraphs carry the message alone | Read only the headings and first sentences aloud. Does the pitch still make sense? |
| Satisficing | The first plausible link is the right one | Ask a colleague to find your price or booking page. Note the first thing they click |
| Conventions | Logo links home, navigation sits where people expect it, links look clickable | Click the logo from three inner pages. Hover over body text to spot links that don't look like links |
| Mindless choices | Menu labels are your customers' words, not internal names | List your menu labels. Cross out any a customer wouldn't say out loud on the phone |
| Omit needless words | No welcome text, no instructions nobody reads | Delete the first paragraph of your homepage in a copy. If nothing is lost, delete it for real |
| Street signs (trunk test) | Every page shows the site name, the page name, the main sections and where you are | Open a random inner page on your phone. Can you name the site, the page and how to get back? |
| Homepage first questions | What you do, for whom and where to start, above the fold on mobile | Screenshot the first mobile screen. Circle the answer to each question. Any blank is a fix |
| Goodwill | Price ranges, phone number and contact options are easy to find | Time how long it takes to find your phone number and a price signal. Over 10 seconds is too long |
| Accessibility | Text contrast, alt text, keyboard use | Tab through the page without a mouse. Can you reach and see every link and button? |
Write down every place where someone hesitated. Hesitation is the finding. You don't need a score.
What changes in 2026 when AI answers come before the click?
Our view: these rules matter more in 2026. More visitors now arrive mid-decision, and they land on a deep page instead of the homepage. When an AI tool summarizes the options first, the visitor who clicks through has already read a paragraph about you. They want to confirm one fact and act.
The click itself is scarcer. A Pew Research Center study tracked the browsing of 900 American adults. When Google showed an AI summary, people clicked a regular result on 8% of visits. Without a summary, they clicked on 15%. Clicks on links inside the summary itself happened on about 1% of visits.
For a business site, three things follow:
- Every page is a homepage. The AI answer often links to a service page, a pricing page or a blog post. That page has to pass the trunk test and answer "what is this, and where do I go next" on its own.
- The visitor's question is narrower. Someone who asked an AI "who does accessible websites in Toronto for clinics" arrives wanting proof of that exact thing. A page that buries it under a generic intro makes them think, so they go back to the answer.
- Words matter twice. Plain headings and short, factual sentences help a scanning human. They are also the lines an AI engine can lift and quote. So "omit needless words" helps answer engine optimization too, meaning getting named inside AI answers.
Adding a chat box doesn't fix a confusing site. An empty prompt field asks visitors to do the thinking, which breaks the first law. We wrote about that in The Blank Text Box Is a Trap.
If you're planning a redesign, put these checks in the brief. Our list of questions to ask a web designer before hiring covers what to get in writing. We also explain why we think a page structure built for your own customers beats a stock layout in why custom web design beats templates.
How do you run usability testing on a small budget?
Krug's answer is to test with three people, one morning a month, and debrief over lunch. He described the routine in a 2011 interview with .net magazine. The team watches the sessions, then decides over lunch which problems are worst and what to fix before the next round.
He admits three people is not statistically valid. His point is that watching three people still turns up many of the most serious problems. Doing it yourself, he said, costs a couple of hundred dollars. A hired round costs between five and ten thousand dollars. He also said to start testing early in the design, not once near the end.
The part teams skip is the watching. Krug was blunt about it in that interview.
> "…get as many people on the team as possible to come and observe." > Steve Krug, Creative Bloq / .net magazine, 2011
His companion book, Rocket Surgery Made Easy, turns the routine into a method. The downloads page on his site gives away a test script for websites, a recording consent form and planning checklists. The script tells the participant that the site is being tested, not them, and asks them to think out loud. The checklists plan a session from three weeks out to the debrief lunch.
Here is the small-business version we suggest.
- Who: three people who aren't on your team and don't know your business well.
- What: two or three tasks a real customer does, such as "find out if they serve Mississauga" or "book a consultation."
- How: a laptop or a phone, a screen recorder and Krug's script, read word for word.
- Then: one hour, everyone who watched, one list of the three worst problems. Fix those before testing again.
Three people won't tell you which headline converts better; that needs traffic and an A/B test. Testing friends who already know your business hides the confusion a stranger would feel. And a test only helps if someone has the authority to fix what it finds.
Next step
Want an outside read on your own pages? Our free site check at /audit reviews how your site reads to people and to AI answer engines.