Multi-page metadata comparison

How to use it

1. 1Paste the addresses, one per lineUp to twenty, and they may be on different sites — that is the point of this tool rather than the crawler. Comparing your page against three competitors is exactly what it is for.
2. 2Press Compare and waitEach address is fetched in turn, politely. robots.txt is obeyed even though you typed each address by hand, because twenty requests from one submission is the request volume that rule exists to guard against.
3. 3Read the table down the columnsTitles, descriptions and headings are lined up so the odd one out is visible at a glance. That is the whole value of the tool — the same three facts about five pages, side by side.
4. 4Check the rows that could not be readAn address that failed stays in the table with its reason and is excluded from every count. One refusal does not fail the batch, and a page that could not be fetched is never quietly dropped.
5. 5Treat a shared title across different sites carefullyTwo competitors using the same title is information, not a fault, so it is reported as a note rather than a problem. Duplicate titles matter WITHIN a site.
6. 6Share or keep the result within 24 hoursThe comparison has its own link. It is deleted 24 hours after it ran, so save it if you need it for a report.

About this tool

Paste the addresses you care about and see their metadata in one table: title, meta description, H1, canonical, Open Graph tags and word count, side by side. Up to twenty pages, and they do not have to be on the same site.

That last part is what makes this different from the duplicate/thin detector, which crawls a site and reports what it finds inside its limits. Here you choose every address — pages a crawl would never reach, a set you are about to publish, or your page next to a competitor's.

Understanding the result

Read the count at the top first. Everything below is out of the pages we could actually read, never out of the pages you submitted — an address we could not fetch is not evidence of anything.

A page we could not read stays in the table with its reason instead of disappearing. A missing row gets read as "this page has no metadata", which is a claim about your page rather than about our failure to reach it.

When the addresses span more than one site, shared values are marked as such and their severity drops. Two of your own pages sharing a title is a problem you can fix; your page sharing one with somebody else's is not the same thing, and reporting them the same way would be advice you cannot act on.

A page whose field is absent is not counted as sharing a value with another page that is also missing it. "Three pages have no description" is a separate finding, and merging the two would describe a set of unfinished pages as heavily duplicated.

Example

https://example.com/products/a
https://example.com/products/b
https://example.com/products/c

Compared 3 of 3 addresses

critical  3 pages share a title — "Products | Example"
warning   2 pages have no meta description

/products/a   Products | Example   —              Our products   self
/products/b   Products | Example   —              Our products   self
/products/c   Products | Example   Buy widget C   Widget C       self

Limitations

• Only the addresses you give are fetched. Nothing is followed, nothing is discovered, and a page you did not list is not considered — if you want that, the full-site crawler is the tool.
• robots.txt is obeyed for every address, and in bulk mode: several requests from a single submission is bulk fetching whoever typed them, so a site asking crawlers to stay out is respected here. To check one page on such a site, use the single-page metadata checker instead.
• Pages are read as delivered, without running JavaScript. Metadata written by a script in the browser will not appear in this table.
• There is no indexability verdict. That needs each page's robots.txt rules, which would be extra requests, and re-deriving it here would mean a second copy of a rule the indexability checker already owns.
• Twenty addresses is the maximum. It is twenty requests to somebody's server, made one at a time with a pause between them, and a tool that accepted a thousand would be a load generator with a form on it.
• Results are deleted after 24 hours. Requests to private and internal addresses are refused by design.

Questions