How do you find orphan pages on a website?
Finding orphans is a set subtraction, and understanding it that way explains why a crawler alone can never do it. A crawler discovers pages by following links. An orphan has no incoming links. The crawler will therefore never reach it, which means the pages missing from your crawl are exactly the ones you are looking for.
The subtraction that finds them
You need two lists. The first is every URL that exists on the site, which cannot come from a crawl. The second is every URL that a crawl actually reached by following internal links. Orphans are the URLs in the first list and not the second.
The whole method in one lineorphans = (all URLs that exist) - (URLs reached by following internal links)
Everything else is detail about where to get the first list.
Sourcing the list of URLs that exist
No single source is complete, which is why serious orphan hunts combine several. Your XML sitemap is the obvious start, but it only contains what your sitemap generator knew about, and pages missing from the sitemap are often the same pages missing from your link graph. A CMS or database export is usually the most complete inventory, because it lists what was published regardless of whether anything ever linked to it.
Server access logs add the pages that real users and crawlers actually requested, which surfaces old URLs that still receive traffic from external links or bookmarks long after your navigation stopped mentioning them. Search Console adds pages Google has indexed, and a page Google indexed but your crawl never reached is an orphan by definition and a particularly valuable one to find, because it is already earning impressions with no internal support.
Reading the results honestly
The raw output of the subtraction always overstates the problem, and working through the false positives is most of the job. Pages reachable only through a form submission or a search box are not orphans in the harmful sense: they were never meant to be crawled. Pages behind a login are correctly outside the link graph. Paginated URLs, filtered variants and print versions often appear because the sitemap lists them while navigation does not, and the right fix there is usually to remove them from the sitemap rather than to link to them.
What remains after that filtering is the real list, and it is usually shorter and more valuable than the raw export suggested. Judge each survivor by whether you want it ranking. A page you want ranking needs contextual internal links from topically related pages. A page you do not want ranking should be noindexed or removed rather than linked to, because adding links to a page that should not exist just spends link equity on it.
Why this is worth repeating on a schedule
Orphans are created by ordinary work rather than by mistakes. A category gets restructured and the pages it used to list lose their only inbound link. A seasonal landing page drops out of the navigation when the campaign ends. A template change removes a related-articles block that was quietly carrying the link graph for a whole section. None of these look like errors at the time, and none of them announce themselves afterwards.
Running the subtraction after any structural change catches these while the context is still fresh, which is considerably cheaper than discovering a year later that a directory of pages has been invisible since a redesign.
Kalenux