Compare Two Pages of Your Site Side by Side

Compare with another page reads two pages of one of your sites right now and shows what differs between them. It compares the status, whether each can be indexed, the title, meta description, canonical, robots meta tag, X-Robots-Tag header, language, H1 and H2 headings, the types of structured data (not their values), counts, and the main text, word by word. Use it to see why two pages look alike to a search engine, what changed between a page and its replacement, or whether a redirect lands where you think.

Open a page from Site Crawler, then Site Pages, press its address (in the Pages view you can also press Details), and use Compare with another page in the page's header. A comparison sends requests to your site, so viewers of a workspace do not see the button. Only HTML pages of the site can be compared: on a file, an image or a link to another site the button is greyed out, and its tooltip says why.

How it works

The page you opened is page A. Search for page B by part of its address, then press it in the list: Compare pages stays greyed out until you do. The list shows the first 50 matching pages of the same site's latest scan, each with its status and title. It has no next page, so type more of the address to narrow it down. A page the last scan could not read is marked.
Press Compare pages. PerfBee reads both pages live, not from the last scan, and shows them side by side. This usually takes a few seconds and can take up to a minute. The fields that differ are highlighted, and Show only the differences hides the ones that are the same.

What PerfBee sends to your site

Usually three requests: one for each page and one for the site's robots.txt. Nothing else: no images, scripts or other files. Nothing is retried. The robots.txt is read once for each address a page is requested from: http or https, www. or bare, and the target of a redirect. A redirect of robots.txt itself is followed too. So a comparison can send more than three requests, but never more than 20.
The requests come from PerfBeeBot, from the same address as a scan, at the pace your site's robots.txt asks for (Crawl-delay or Request-rate, up to 10 seconds between requests). A page that robots.txt does not allow PerfBeeBot to read is not requested, and the comparison says so. If robots.txt cannot be read (your server answers 5xx or 429, or does not answer), PerfBeeBot stays away. A robots.txt that is missing or refused (404, 403) counts as no rules, and the pages are read.
A page that redirects inside your site is followed, up to 3 redirects, and each one is a request. A redirect to another site is not followed.
If your site answers 429 or 503, which means "slow down", the other page is not requested. A scan of the same site waits for the pause as well.

Only pages of the site in your inventory can be compared, from the folder the site covers. You cannot type an address.

Reading the result

Each field is Same, Changed, Only on A or Only on B. For lists such as headings and structured data types, the entries only one page has are marked.
Words are shown in three rows. Words (as the crawl counts) is the number a Detailed scan stores for the page. It counts the text of the page's first main element, else its first article, else the whole body (the menu and footer then count too), including the text of scripts and styles. Words in the visible text is every word a visitor can read, menu and footer included. Words in the main content leaves out the menu, footer, sidebars, forms and dialogs. The text comparison reads the main content.
Main text says what share of the words match and shows both pages' text in two columns. Red, struck-through words are only on A, green words are only on B, and long stretches that did not change are folded into a line such as "… 120 unchanged words …".
A page that answered an error such as 404, was kept out by robots.txt, is not HTML or could not be reached is not read as if it were the page. When either page is one of these, only the HTTP status of the two pages is compared (a page that gave no answer shows none), and a note says why. No other field and no text is compared, and the Show only the differences box does not appear.

Limits

PerfBee reads the HTML your site sends, as a scan without JavaScript rendering does. Content a script adds after the page loads is not in the comparison.
A page is read up to 2 MB, and the first 3,000 words of each page's main content are compared. When a page is cut, the result says so.
A page behind a login, or on a site that lets the crawler in only with a token or a staging password, shows the answer a visitor without them would get, such as 403.
A workspace can run 100 comparisons a day. The day is counted in UTC and starts again at midnight UTC. One network address can start 6 comparisons a minute.
A comparison counts toward the 100 once it starts, also when a page could not be read. When several comparisons are already running, PerfBee asks you to try again in a moment: that try does not count.

Cost and what is kept

A comparison uses no credits. PerfBee does not keep the pages, their text or the result: close the window and it is gone, and the inventory is not changed. Use Recheck on a page's detail to update the page's own row.