The short answer
SharePoint search is usually not broken, it is bounded. Microsoft publishes hard limits on what its crawler will index: parsing stops after 2 million characters per item, only the first 1,000,000 characters are tokenized, files over 150 MB (512 MB for PDF, PPTX, PPT, DOC and DOCX) have only their metadata indexed, and any item accessible to more than 10,000 distinct users or security groups "isn't searchable by any user." Add the settings that silently exclude a site or library from results, the fact that schema changes do not trigger a reindex, and the reality that SharePoint only sees Microsoft 365 content, and most complaints about SharePoint search trace to one of seven specific causes rather than to bad relevance ranking.
Last updated: August 2026. Sources: Microsoft Learn, Search limits for SharePoint; Manually request crawling and reindexing of a site, a library or a list; Search results missing in SharePoint Online.
"Why is SharePoint search so bad" is one of the most searched questions about Microsoft 365, and the usual answers are unsatisfying. Blog posts tell you to add metadata, or write better titles, or train people on refiners. Sometimes that helps. More often the document a person is hunting for was never fully indexed in the first place, and no amount of query skill will surface content that is not in the index.
The useful thing about SharePoint is that Microsoft documents its search boundaries precisely. So instead of guessing, you can check your situation against the published limits and find out whether you have a configuration problem, a content problem, or a scope problem. Those three need completely different fixes.
How does SharePoint search actually work?
SharePoint crawls your sites on an automatic schedule, picks up content and metadata as crawled properties, and maps those to managed properties. Microsoft's wording matters here: "Only managed properties are kept in the index. This means that users can only search on managed properties." Whatever does not survive that pipeline does not exist as far as search is concerned.
The crawler is incremental. It picks up what changed since last time. That is efficient, and it is also the source of a whole category of confusion, because changes you make to the search schema are not changes to the site, so the crawler does not notice them at all.
The limits Microsoft publishes
These come straight from Microsoft's search limits reference. They are worth reading once, because several of them explain complaints that otherwise look like random flakiness.
| Limit | Maximum | What it means in practice |
|---|---|---|
| File size the crawler will download | 150 MB all types; 512 MB for PDF, PPTX, PPT, DOC, DOCX | Above the limit, "only document metadata is downloaded, while the full content is NOT downloaded and remains unavailable for search." |
| Parsed content size | 2 million characters | Parsing stops there, and also stops after 30 seconds on a single item. The item is marked partially processed and the rest is never indexed. |
| Characters the word breaker processes | 1,000,000 | Only the first million characters become searchable tokens, again subject to a 30 second ceiling. |
| Distinct users or groups with access to an item | 10,000 | Past that, "the item isn't searchable by any user." The most widely shared documents can be the least findable ones. |
| Unique indexed tokens per managed property | 10,000 | Beyond it the index keeps the first 10,000 and flags the file as partially processed. |
| Rows in a result set | 500 | You page through more, but a single query will not hand you a complete large set. |
| Keyword Query Language text length | 4 KB (4,096 characters) | Relevant if you generate queries programmatically rather than typing them. |
Seven real reasons SharePoint search is bad
1. The site or library is switched off for search
This is the first thing Microsoft support tells you to check, and it is the single most common cause of "the document exists but search cannot find it." Two separate settings control it. At site level, Allow this site to appear in Search results must be set to Yes, under Site Settings, then Search and offline availability. At library level, Allow items from this document library to appear in search results? must also be Yes, under Library Settings, then Advanced Settings.
Either one set to No removes content from results for everyone while leaving the files perfectly accessible by navigation. That mismatch is why it goes undiagnosed for months: people can open the document from a link, so nobody suspects indexing.
2. Schema changes do not reindex themselves
If someone added or remapped a managed property and the new field still returns nothing, this is why. Microsoft is explicit: "Because your changes are made in the search schema, and not to the actual site, the crawler won't automatically reindex the site." You have to request a reindex of the site, library or list, and the content is then picked up on the next scheduled crawl.
Microsoft also attaches a warning to the site-level version: "Reindexing a site can cause a massive load on the search system. Don't reindex your site unless you've made changes that require all items to be reindexed." Reindex the specific library where you can.
3. Large files are indexed by name only
A 700 MB PDF of the archived engineering handbook is in SharePoint. Its title is searchable. Its contents are not, because the crawler never downloaded them. Anything past 150 MB, or 512 MB for PDF and the Office formats, contributes metadata and nothing else.
This one is easy to test. Search for a distinctive phrase you know appears on page 300 of the file. If the file comes back when you search its title but not when you search that phrase, you have found your answer.
4. Long documents are truncated
Even under the size limit, parsing stops at 2 million characters, and word breaking stops at 1,000,000. A long compliance manual, a merged contract pack or a spreadsheet with a lot of text can quietly cross that line. The item is marked partially processed and the tail is invisible. In practice, the deeper into a long document a fact sits, the less likely search is to find it, which is the opposite of what people assume.
5. The most-shared documents can be unsearchable
This is the limit almost nobody knows, and it is worth quoting exactly: "When more than 10,000 distinct users or Active Directory / Microsoft Entra security groups have access to an item, the item isn't searchable by any user."
Read that again if you work at a company with more than 10,000 people. The all-hands deck, the employee handbook, the security policy, the exact documents you most want everyone to find, are the ones most likely to be shared broadly enough to trip this. Microsoft notes such items are still returned by eDiscovery queries, which is little comfort to an employee looking for the travel policy.
6. Broken permission inheritance hides pages
Microsoft's troubleshooting guidance covers a subtler one. Site search visibility has three options, and the default is "Do not index Web Parts if this site contains fine-grained permissions." The consequence: "If a subsite on the site collection doesn't inherit permissions, .aspx pages won't appear in the search results."
Any team that broke inheritance on a subsite to lock down one library may have removed all of that subsite's pages from search as a side effect. The fix is to set the option to "Always index all Web Parts on this site," or restore permission inheritance.
7. It only sees Microsoft 365
The six causes above are fixable by an administrator. This one is architectural. SharePoint search indexes SharePoint, and Microsoft Search extends that across Microsoft 365. Neither reaches Slack, Notion, Google Drive, Jira, Salesforce or the tools your teams actually decide things in.
That matters because of how workplace knowledge is really shaped. The policy is a SharePoint page. The exception to it was granted in a Slack thread. The ticket that implemented the exception is in Jira, and the customer commitment that triggered it is a note on a Salesforce opportunity. Search the intranet and you get a quarter of the answer, correctly, with good permission handling, and it is still the wrong answer. If the real question is where a reported number came from rather than which document mentions it, that is a data lineage problem you trace back through the warehouse, not something any document search can settle.
Microsoft Search vs SharePoint search: what is the difference?
People use the terms interchangeably and they are not the same thing. Classic SharePoint search is scoped to the site or site collection you are standing in, which is why searching from a team site and searching from the tenant root give different results for the same words. Microsoft Search is the tenant-wide experience in the Microsoft 365 header, covering SharePoint, OneDrive, Teams, Outlook and people.
Microsoft Search is broader and generally the better starting point. It still indexes only what Microsoft 365 holds, and it still inherits every limit in the table above, because the index underneath is the same one.
How to search SharePoint effectively today
While you work through the causes, these help with what you already have.
- Search from the right altitude. Searching inside a site scopes to that site. Use the Microsoft 365 header search for anything tenant-wide.
- Use property filters. Keyword Query Language supports
filetype:docx,author:,path:andlastmodifiedtime. Combining a filetype with two content words beats a long natural phrase. - Search for phrases that would appear in the body, not the title. Titles are often generic. A distinctive sentence from inside the document is a stronger signal.
- Check whether the file is a scan. An image-only PDF with no text layer contains nothing to index. It will never be findable by its contents no matter what you configure.
- Verify a missing document is indexed at all before blaming ranking. Search its exact filename. If that fails, it is an indexing problem, not a relevance one.
When configuration is not the answer
Work through the seven causes honestly and you will usually land in one of two places.
If the problems were settings, crawls and oversized files, fix them and SharePoint search gets meaningfully better. It is a competent search engine with genuinely good permission handling, and a well-configured tenant is a reasonable place for Microsoft-centric knowledge to live.
If you fix all of it and people still cannot find answers, the constraint is scope, and no setting in the SharePoint admin center addresses scope. At that point the question stops being "how do I improve SharePoint search" and becomes "how do we search everywhere at once." That is what an intranet search engine that reads past the intranet is for, and it is the same job covered from the Microsoft side on our SharePoint search tool page and more broadly under searching across all company apps.
The bar to hold any replacement to is not "does it have AI." It is whether it indexes the systems SharePoint cannot reach, whether it returns a cited answer instead of another results page, and whether it enforces item-level permissions per person at query time so that a company-wide rollout does not widen access to anything. Microsoft gets that last part right, and anything you put beside it has to as well. Teams weighing the native option first usually start with the Microsoft Copilot alternative comparison and the licensing math behind SharePoint agents.
Frequently asked questions
Why is SharePoint search not working?
Check the two visibility settings first: "Allow this site to appear in Search results" at site level and "Allow items from this document library to appear in search results?" at library level. If both are Yes, the next most likely causes are a schema change that was never reindexed, a file above the crawler's download limit, or a document shared with more than 10,000 users or groups.
Why can I not search in SharePoint?
If the search box itself is missing rather than returning nothing, the page layout or a site template change has removed it, and a site owner can restore it. If the box is present but results are empty for content you can open by navigating, that is an indexing problem, and the site or library search visibility setting is the place to start.
How long does SharePoint take to index new content?
Microsoft does not publish a fixed interval. Content is crawled "automatically based on a defined crawl schedule," with the crawler picking up what changed since last time. New content typically appears within minutes to a few hours. If you have changed the search schema rather than the content, it will not appear at all until you request a reindex.
Does SharePoint search index PDF contents?
Yes, for text-based PDFs up to 512 MB, which is the extended crawl limit for PDF along with PPTX, PPT, DOC and DOCX. Two things break it: files above that size, where only metadata is indexed, and scanned or image-only PDFs, which hold no text layer for the crawler to extract.
Is Microsoft Search better than SharePoint search?
For most people, yes, because it is tenant-wide rather than scoped to the site you happen to be in. It runs on the same index and inherits the same limits, so it fixes scope inside Microsoft 365 without changing what gets indexed or how much of each file is read.
Can AI fix SharePoint search?
Only partly, and it is important to be precise about why. AI improves how a question is interpreted and how an answer is written. It does not change what is in the index. If a document was truncated at 2 million characters or never crawled, an AI layer over the same index still cannot cite it. AI helps most when it is paired with wider indexing across the apps SharePoint does not reach.