Indexing errors divide into two stages, and knowing which stage failed tells you who has to fix it. Download errors mean the robot never got the document. Processing errors mean it got it and could not use it.
Download errors and their limits
- Text size limit exceeded — the document is over 10 MB.
- Maximum URL length exceeded — over 1,024 bytes in the domain name or in the host page address.
- Maximum HTTP header length exceeded — Yandex names the usual cause outright: too many cookies. A tag manager or consent platform stacking cookies can make pages unfetchable while every browser copes fine.
- Invalid message length — check
content-length; the declared header size must not exceed the actual document size. - Invalid data amount transferred — transfer ended early or continued past the declared length.
- Wrong encoding —
Transfer-Encodingset incorrectly or unknown. - Plus the connectivity family: connection failure, connection not established, DNS error, invalid HTTP status code, invalid HTTP header, unsupported document format, invalid document address, and blocked in
robots.txt.
Processing errors
noindexmeta tag present — deliberate, if you meant it.- Wrong encoding — characters that do not match the declared encoding.
- Document recognised as a server log — an unusual one, and Yandex’s advice is telling: if it is a real page for visitors, change it so it does not resemble a log. Pages that are mostly timestamped rows can fall into this.
- Invalid document format — the file does not match its declared type across HTML, PDF, DOC, RTF, SWF, XLS and PPT.
- Encoding not recognised — possibly no text, or text in mixed encodings.
- Language not supported — the language was not recognised or is not supported.
- Document contains no text — the case behind a large share of “the page looks fine to me” reports.
How to use this
Read the error, decide which stage it belongs to, and route it accordingly: download errors go to whoever runs the server or the CDN; processing errors go to whoever builds the page. Sending the wrong half to the wrong team is the most common reason these sit unresolved for months.
How we apply this
The two that surprise clients are the header-length limit caused by cookie accumulation and “document contains no text” on a page that renders perfectly in a browser. Both are invisible without checking what the robot actually received. We reproduce every one of these with the server response check before writing a single ticket, because the fix depends entirely on which stage failed.
Related services