“Short + clear + with keywords + shallow hierarchy + stable without changes” is the URL naming method that best complies with Google E-E-A-T and SEO standards. For example: /seo-url-structure-guide, which can increase click-through rate by approximately 20%-30% compared to random parameters (such as ?id=123). At the same time, the URL hierarchy should be controlled within 3 levels or less (such as /blog/seo/url), which helps search engine crawling and weight distribution.
Must Include Core Keywords
“Good URLs” vs “Bad URLs”
Ordinary website backends store many numbers. Pages with /item-45992 only receive 3 visits in a month. After changing to /sony-headphones, crawlers visit to crawl 14 times per week. WordPress website building tools tend to generate addresses in the style of /?p=204.
Ahrefs organization randomly sampled 2 million search terms. Pages with long strings of pure numbers rank beyond the 30th position. Short English words allow machines to understand the product within 0.02 milliseconds. Deeply nested /category/electronics/product-99 occupies 32 character positions.
Desktop pages forcibly truncate URLs exceeding 70 characters. Microsoft’s eye-tracking device recorded visitors’ gaze停留 movements. Everyone spends 0.8 seconds scanning the green URL area before clicking on a webpage. Readable words increased the click-through rate by 25%.
| What the URL Looks Like | How Many Characters | How Long Machine Takes to Read | Do Users Want to Click |
|---|---|---|---|
| /?id=99283&sort=price | 21 | 0.09 milliseconds | Extremely low |
| /c/apparel/shirts/blue/item22 | 29 | 0.07 milliseconds | Relatively low |
| /mens-blue-cotton-shirt | 23 | 0.02 milliseconds | Extremely high |
Encoding links with a bunch of %E4%BD causes 18% of people to give up clicking. Google engineers mentioned the usage of hyphens. /black-running-shoes is split into three clean words. Underscores make black_running_shoes a weird word composed of 19 letters.
Stuffing 4 similar product names into one address causes trouble. The page’s ranking plummets within days. Keeping only 3 to 5 descriptive words for things is the safest approach. A website selling outdoor tents rewrote links for 4,000 products.
Abandoned /catalog/item?id=44 replaced with /waterproof-hiking-tents. After 45 days, the backend gained 130% more free visitors. A few actions saved a lot of trouble:
- Remove the, a, an, freeing up 6 to 9 character display space.
- Type all letters in lowercase, Linux systems treat uppercase and lowercase as two different pages, easily causing 404 errors.
- Slashes
/should not exceed two; each additional symbol deducts 15% of the weight.
Backlinko reviewed 11.8 million records. URLs ranking in the top three average only 50 characters. Links ranking ninth and tenth average 62 characters. Non-English characters become 45 percent signs plus number codes when copied.
Twitter platform automatically truncates overly long links. Shopify system likes to stuff things into /products/. Redundant /collections/all/products/ creates two addresses for the same product. Backend tools report errors, and traffic on both sides gets pulled down.
Adding year to /2023/best-laptops becomes old news after 12 months. Updated the article, but 2023 still hangs in the link. Removing the time tag keeps articles fresh for years. Old numeric URLs need proper 301 redirect code.
90% of accumulated weight can be safely transferred to the new English address. Rewriting the .htaccess file in the server to establish the correspondence between old and new addresses.
Safely Embed Keywords
Backend error prompts occasionally pop up red background white text warnings. Stuffing the same product English name 5 times consecutively into a webpage’s address bar, the algorithmic bot will tag this page as cheating within 48 hours. The first step to safely filling in words is to control the frequency. Put only 1 main English word in a link.
Moz research team reviewed 10,000 product selling pages. They found that target English words positioned in the first 5 characters after the domain suffix can gain 22% more crawler capture share. Writing domain.com/black-shoes is much better than the extended version domain.com/store/category/items/black-shoes. Machines recognize the first few letters in less than a millisecond.
Making room for the main English word requires trimming extra characters. Common browsers’ address bars can usually fully display 70 to 75 English letters. Characters beyond that are forcibly hidden by three small black dots. Casually removing a few short words while typing saves visible space.
- Remove “in” to free up 2 character spaces
- Delete “the” to open up 3 blank spaces
- Remove “and” to save 3 crawl bytes
- Replace all spaces with hyphens –
- Avoid & symbols to prevent page truncation
Spaces in webpage URLs are forcibly translated into %20 – these three characters in the computer backend. Typing “iphone 15” in a URL becomes “iphone%2015”. Processing these three extra special symbols costs the server engine an additional 0.05 milliseconds. Sites with daily visits exceeding 100,000 burn through hundreds of megabytes of bandwidth every day for no reason.
Too many slashes / dilute the English word you want to display. Each additional delimiter slash drops the crawl score of words positioned after it by 10% to 15%. Some bloggers habitually stuff articles into /blog/2024/news/tech/ folders with 4 deeply nested levels.
Crawlers don’t see what that diary entry is about until reaching the 45th character. Removing two layers of extra category directories moves the word you want machines to see 15 characters forward. Online store category pages with sizes and colors easily accumulate a bunch of useless parameter codes unnoticed.
Handling thousands of product categories with color sizes requires some physical isolation actions.
- Block trailing parameters with ?size=XL and question marks
- Add nofollow to &color=red to disallow crawling
- Attach a Canonical tag to the original product page as an anti-counterfeit seal
- Block the entire /filter/ directory in robots.txt
- Let external search crawlers only read the cleanest product name
In the algorithmic bot’s dictionary, singular and plural forms look identical. Writing “shoe” or “shoes” both go into the same matching word database. There’s no need to stuff both similar-looking words into one address just to complete English grammar. Stuffing two synonyms invites Spam detection risk.
Search Engine Journal media investigated 300 websites with traffic drops. 68% of cases happened when the newly redesigned webpage didn’t build a bridge. Visitors from old URLs accumulated over 3 years clicked in only to see a 404 page not found white screen. Search scores accumulated over three years zeroed out overnight.
Renaming a webpage requires writing a 301 permanent redirect code in the server file. Set up redirects so old links transfer 85% to 90% of past scores to new pages within 14 to 21 days. Servers using Nginx environment can complete the transfer by adding two short lines of code in the .conf file.
Links with @ or # or $ symbols have a 27% probability of being rejected by search engines. Addresses spelled with weird symbols make crawlers wander in dead ends. Machines distribute only 300 to 500 page crawl quotas to new websites daily. Wasting server quota on garbled code is extremely costly.
Double-check a few details before saving new content.
- Use standard English lowercase letters when typing
- Word count should stay between 3 to 5 simple English words
- Carefully check for any mixed uppercase letters
- Review if any extra commas or periods are left behind
- Absolutely no meaningless Pinyin initials
Separate Words with “-” (Do Not Use Underscores)
Google Crawler’s “Reading Habits”
Human eyes see things extremely fast. The visual nerve transmits a string of letters to the visual cortex in less than 250 milliseconds. An ordinary reader scanning newyorktimes, the visual system immediately retrieves memory and interprets it as three familiar English words.
Machines see things according to a completely different set of rules. The server hosting webpage content sends a string of URL code, and crawlers can only recognize symbols one by one according to the most basic encoding table.
- Position 45 in the encoding table is the hyphen
- When crawlers read it, they insert a space
- Position 95 in the encoding table is the underscore
- Machines treat it as part of an English word
- The program won’t re-split connected letters
A server located in an Oregon data center sees hundreds of billions of webpages daily. Time allowed for recognizing a single webpage is firmly capped within 15 milliseconds. When the time limit expires, unrecognized words are果断 dropped by the machine.
Adding a single hyphen in the URL makes crawling effortless. Machines package the split independent words and load them into an index database containing 13 billion webpage records. When people type in search boxes, algorithms search the database for matching words and retrieve related pages at millisecond speed.
Underscores ruin the entire recognition process. When crawlers receive black_leather_jacket, the underlying C++ machine code kicks in, and the program treats this string of connected letters as an entirely new ultra-long word.
- Spell-checker reports errors and can’t find this word
- Crawler follows instructions to forcibly find similar letter replacements
- Long concatenated words are thrown into backend error categories
- Original high-frequency word perfect matches completely fail
- Page enters a 30-day low-level waiting pool
Checking backend real-time logs is extremely straightforward. An article with three thousand words, URL with underscores, after six weeks online, organic crawl times on the search console is only a meager 7 times.
Reviewing a digital analysis of 50,000 dead link pages, URLs connecting four or more words with underscores have a 73% machine rejection rate for inclusion. Made-up words not found in the machine dictionary forcibly block pages from entering the search database.
Replacing all underscores in URLs with hyphens and resubmitting the sitemap. The same article saw crawl requests surge to 120 times within two days. Following machine rules and changing one symbol eliminates millions of useless guessing calculations.
Human eyes are extremely sensitive to character spacing on screen. Numbers from eye-tracking tests show that once letter spacing falls below 0.5 millimeters, eye gaze stay time doubles. When people scroll on phone screens, the time spent examining a single link is only 0.4 seconds.
- Hyphens visually create appropriate pauses
- Reading burden on small screens is reduced for long URLs
- Audiences casually sharing links aren’t easily treated as garbled text
- Overseas social media external link clicks recover
Programmer Mindset Impact
From the day of typing the first line of code, underscores become muscle memory for programmers. Looking at statistics from GitHub, the top 3 open source code repositories, 87% of Python projects name variables with underscores. People who code press the underscore key approximately 400 times per day in software.
With eyes fixed on black screens, underscores give compact English letters a 1.2 millimeter gap. Machines reading code don’t report errors with underscores. Computer RAM completely absorbs this long string of letter combinations.
A development team took over building a shopping website selling 50,000 clothes. The backend developer wrote a batch URL generation tool, casually replaced all spaces in product names with underscores. Less than three days after the website launched, 32,000 product pages got stuck in the search engine’s waiting queue.
Crawlers reading the large shopping site’s XML sitemap. When encountering underscore-connected URLs, server feedback read time surged from 25 milliseconds to 180 milliseconds. Exceeding the 100 millisecond quota, the crawler program immediately disconnects and moves elsewhere.
| Business Scenario | Common Practice | Letter Format | Search Engine Response |
|---|---|---|---|
| Software Code | Snake Case Naming | user_login_page | Spell Check Error |
| Backend Charts | Underscore Connect | product_detail_info | Cannot Split Words |
| Webpage URLs | Hyphens | winter-warm-jacket | Accurate Indexing |
Flipping through server log files of 200 North American startup sales companies. Sales websites launched under pure technical personnel leadership, over 90% of URL directories carry coding traces. What the backend database chart is called, the front-end URL copies exactly.
Technical staff find things extremely fast by typing underscores in the database. Push 5,000 product entries from the database to the front-end web page. Developers wrote two fewer lines of conversion code, and underscore long URLs were completely exposed to search engines.
- Enter English original names with spaces into the backend
- Code forcibly converts spaces to underscores stored in database
- Front-end server generates product webpage links as-is
- Crawlers process as continuous long words and reject
- 20,000 pages on the search report show impressions dropping to single digits
Numbers from the backend data panel have no warmth. An outdoor camping equipment site, 4,000 sleeping bag product pages online for six full months. Search click volume on the report is less than fifty times. People typing “waterproof sleeping bag” in the search box simply cannot see the page written as waterproof_sleeping_bag.
Typing a few lines of replacement code, the entire site’s over 10,000 URLs replaced symbols in one batch. Hundreds of thousands of lines of code ran in less than a minute. 503 server timeout errors received by the search console dropped 85% within two weeks.
The machine’s crawl quota was re-released. In the second month after replacing underscores, daily page reads by crawlers surged from 500 to 4,500. Product descriptions hidden in the underlying database became high-scoring pages generating 100,000 monthly impressions.
Programming language syntax is extremely strict. C language absolutely does not allow hyphens in variable names. The system treats hyphens as minus signs for mathematical subtraction. Coders extremely reject hyphens to avoid full screens of red warnings.
Webpage optimization walks a different path. Applying coding habits to URLs costs an extremely high price. A 60-page website health report clearly states that using one wrong punctuation mark costs the website approximately 1.5 million free visits per year.
- Find the code section that generates URLs
- Delete underscore conversion instructions
- Type in five or six lines of hyphen conversion code
- Resubmit the sitemap for 500,000 pages
Hierarchy Should Not Be Too Deep (≤3 Levels)
Crawl Budget & Weight
Google bots are like a time-pressed room checker. Daily dwell time on a single website is calculated entirely in milliseconds. Each additional slash “/” in a URL, the bot’s patience to keep scrolling drops a notch. For URLs with 4 slashes, up to 42% of bots turn around and leave.
Reducing levels to 3 or less saves a lot of read time. Placing pages in domain.com/shoes/sneakers structure, server response completes in under 50 milliseconds. Bots run smoothly, newly written articles enter the search engine database within 15 minutes.
Website homepage tightly holds over 85% of the site’s initial ranking score. Scores flow down layer by layer following the URL slash symbols. Each additional physical directory layer costs approximately 15% of the score along the way. Fewer levels means articles receive more score weight, and rankings naturally rank higher.
Search Engine Journal reviewed millions of webpage archives. URLs buried 5 levels deep, approximately 62% of articles are completely not indexed by Google. Stacking multiple layers of useless folders, the bot’s read timeout pressure multiplies.
| URL Slash Count | Estimated Bot Return Frequency | Homepage Score Retention |
|---|---|---|
| 1 (Category) | Several to dozens of times per day | Landing around 85% |
| 2 (Content) | 2 to 4 times per week | Maintaining around 72% |
| 3 (Long-tail) | 1 to 2 times per month | Falling below the 50% mark |
| 4 or more | Occasionally glancing once every few months | Almost no score received |
Removing one layer of useless category name, new webpages in search results can jump ahead past 2.3 competitors. Google Search Console backend sets a hard standard: pages with crawl times exceeding 2.5 seconds are marked as inferior structure.
Putting years and months in URLs domain.com/2023/11/post is an extremely wasteful practice. Unnecessarily adding two extra server parsing operations. Save the slash-typing effort and let bots spend their scarce time reading article text.
Google engineer John Mueller mentioned in a video conference that articles placed where homepage can be reached in two clicks are extremely safe.
- Pages with 3-level short links, time to first byte shortened by approximately 12 milliseconds.
- Removing one layer of long path, mobile crawler complete HTML rendering success rate surges to 98.5%.
- Shallow URLs under 50 characters, random click probability by users in messaging apps increases by 18%.
Amazon’s millions of products get quickly indexed by the warehouse. The entire site doesn’t use extremely deep physical folder layers, entirely relying on one layer of flat tags supporting tens of millions of daily high-frequency crawls. Ahrefs ran through 2 million popular search terms. URLs ranking in the top 3 positions of the first page average between 48 to 53 characters. Compact URL structure naturally has a quality that algorithms favor.
User Trust (Trust)
When browsing online, the time eyes spend scanning the URL line in search results is only about 1.5 seconds. Microsoft conducted an eye-movement test report, and human eyes’停留 time on a single link is truly extremely short.
Within one and a half seconds, the brain entirely relies on intuition to decide whether to click the mouse. URLs stuffed with 80 English characters and four or five slash symbols, ordinary people see it, their hearts mostly waver. Long strings with random number garble look extremely like phishing pages that trick people into clicking. A U.S. cybersecurity organization statistics show that lengthy URLs with 3 or more levels have a 47% chance of being treated as scam links.
Seeing a 5-level directory long link is absolutely not as reassuring as scanning a 2-level extremely short link. Minimalist URLs instantly establish a strong sense of security.
Minimalist URLs that reassure people have extremely subtle characteristics:
- Character count compressed within 50.
- No messy symbols like
%20or?id=9527. - Words are understandable, no need to guess what’s written inside.
- Fewer than three slashes, files feel shallow.
Today 60% of webpage browsing is done on 4 to 6 inch phone screens. Phone screens are narrow; when search engines display URLs, characters exceeding 65 are definitely forcibly cut off.
URLs too long, the tail becomes three ellipsis points. URLs get truncated, half the characters invisible, ordinary people simply cannot guess what will appear after clicking. Nielsen Norman Group conducted dozens of blind tests. Seeing truncated long URLs on phones, finger tap probability instantly drops 15 percentage points.
Sending URLs to friends via messaging apps is extremely common. A long link with several folder names pasted into a chat window often takes up most of the phone screen. Bitly, a URL shortening service, published tens of millions of sharing records. Clean URLs with minimal levels are clicked by friends in chat apps 34% more times than long links.
The actual action of sending URLs in chat boxes is entirely intuitive:
- The shorter the link, the stronger the willingness to click casually.
- Long links often miss trailing characters when copying.
- Long chains with garble are often treated as mass advertising.
- Short chains go well with
How to Simplify
The commonly used WordPress system for building websites preset stores articles in three levels: year, month, and day. An article’s link unnecessarily gains three deep slashes. Completely removing time numbers from links, search engine bots can read 15% more new pages daily.
URLs with timestamps are easily treated as outdated news by ordinary users. Ahrefs randomly examined 100,000 old articles; after removing year numbers from links, clicks on search results pages increased 22% within three months.
Article content can be updated and rewritten anytime in the backend. The year number in the link is like a tattoo that won’t wash off. Cutting the year number in the link, a three-year-old article looks as fresh as if written today.
Modifying old links must be paired with 301 redirect settings. Hang an action instruction on the server, directing the old deeply nested link to the new two-level extremely short link. Bot following the instruction, the original traffic score accumulated over three years completely intact.
Shopping website category folders easily extend infinitely to 5 levels or more. Menswear, shoes, sneakers layered in. Flatten the link completely, product individual pages all hang under the main domain, webpage loading speed can be 80 milliseconds faster.
Shopify, a standalone site building tool, has a rigid path setting. The system forces all products to be wrapped in the /products/ folder. Rewriting this level in the backend with code, 2,000 product individual pages got back the 18% link weight that was originally eaten.
Navigation menus at the top of pages are for real humans to click. Structure in the URL bar is entirely for algorithms to scan quickly. Category layer count and file actual storage location untied, individual webpage ranking momentum is extremely strong.
Removing middle unnecessary category names:
- Mens shoes category should not write
/men/shoes/, merge to/mens-shoes. - Unpopular product pages skip categories, attach to main domain.
- Remove redundant intermediate directories like blog or article.
- Merge multiple tags into a single link string.
Old websites like to hang .html or .php extensions at the end of URLs. Five extra letters provide absolutely no search information. Removing the suffix and changing to slash ending saves nearly 15 milliseconds in MIME type recognition during page load.
A real data test run on Moz forum. After cutting the redundant /blog/ level, individual article homepage weight increased凭空 by 12%. The weight reflux from simplified paths is extremely powerful.
English domains easily write prepositions into paths. Characters occupy enormous space. Removing a, the, of entirely, a 50 English character quota can hold many high search volume keywords.
An article about fixing broken iPhones, the link doesn’t need to translate into a complete English long sentence. Just keep fix-broken-iphone – three words, search algorithm word matching speed can be approximately 40 milliseconds faster.
| Verbose Link Format | Flattened Link After Compression | Estimated Read Time Saved |
|---|---|---|
/2023/11/05/post-name |
/post-name |
Approximately 35 milliseconds |
/store/geo/us/nyc/shop |
/us-nyc-shop |
Approximately 50 milliseconds |
/category/tech/phones |
/phones-review |
Approximately 42 milliseconds |
Long strings of question marks and equal signs trailing after URLs. Dynamic paths with ?sort=price extremely consume bot crawl quota. Shopping websites turn off indexing for hundreds of thousands of dynamic filter links, bot crawl frequency for high-priced products doubles.
Writing good Canonical syntax directs messy parameter links to one layer of clean short link. Search engines identify the clean URL to distribute scores. Saving tens of thousands of daily crawl calculation wastes.
The habit of storing photo files in computers is easily carried over to website building. Build a year folder, then a month folder, then a scenery photo folder. When search engine bots reach the fourth level, approximately 35% probability disconnects on the spot due to read timeout.
Avoid writing non-English characters into URLs. Browsers automatically encode special characters into extremely long strings of %E6 garble. Encoded link length often exceeds 200 characters, looking very ugly when pasted in chat apps.
Forcing all webpage links to rewrite as English letters plus hyphens. Short pure English minimalist links, when ordinary people scan them at a glance, feel extremely secure. Internet reputation scoring agencies give security ratings an average of 15 points higher.
Tag label pages built into WordPress system easily create disasters. Hundreds of useless tags generate extremely deep invalid links. Clearing domain.com/tag/ underlying fragmented pages, overall site indexing rate climbs nearly 8 percentage points.
News websites are extremely obsessed with building folders by author names. Stuffed hundreds of reporters’ pinyin all into URLs, forcibly adding one layer of physical directory. A certain foreign media digital version already removed hundreds of thousands of author-specific link levels long ago, relying on article page internal links to distinguish author identity.
