Short answer: the backup failed its own consistency check. Your files are probably still there, but Time Machine will no longer trust the history, and there is no repair tool.
Before you accept the offer to start a new backup, get what you need out of the old one. That offer deletes it.
Time Machine keeps a record of what it wrote. Verification re-reads the backup and compares it against that record. When it says the backup cannot be verified, one of the two no longer matches the other: a file is not where the index says it is, or the bytes are not the bytes it wrote.
That is a statement about the index, not usually about your files. Most of the data is still sitting on the disk. What Time Machine has lost confidence in is its ability to walk the history and give you a specific version of a specific file on a specific date, which is the whole point of it.
There is no fsck for a Time Machine backup. Apple ships no repair tool, and this is why
every forum thread ends the same way: delete it and start again.
The dialogue offers to start a fresh backup. Taking that offer removes what you have. Even a backup that failed verification is very likely to hold readable copies of your files, so treat it as the last copy of something until you have proved otherwise.
Backups.backupdb, then your Mac's name, then a dated folder. Those are ordinary folders
full of ordinary files, and you can drag out of them.If the disk itself is behaving badly, clone it before you touch it and work on the clone. Disk Utility, File, New Image, Image from [your disk], and choose read-only. A failing drive gets worse every time it is read, and you get one good attempt at this.
This surprises people who go looking for the setting. Hold Option and click the Time Machine icon in the menu bar and you get Verify Backups, but only when the backup lives on a network share. On a USB or Thunderbolt drive the option does not appear, because macOS does not run that check on local disks at all.
So most people meet this message on a NAS or a Time Capsule, and everyone with a drive on the desk is in the other situation: not verified, and not verified failing either. Nothing is checking.
The closest thing to a manual check on a local disk is:
tmutil compare -swhich compares your Mac against the latest snapshot and lists the differences. It is a comparison, not an integrity check, and on a large disk it takes a long time. It will tell you the backup can be read and walked, which is more than you knew a minute ago.
diskutil verifyVolume /Volumes/YourBackup,
and look at the SMART status in System Information under Storage. A disk with reallocated sectors is
telling you something.Once your files are out, the honest options are short. You can start a fresh backup, which is what Apple's dialogue offers and what almost everybody ends up doing. You lose the history and keep the present. Or you can keep the old disk untouched as a cold copy and start the new backup somewhere else, which costs the price of a second drive and is the better answer if the history mattered.
What you should not do is dismiss the message and carry on. A backup that has failed verification does not recover on its own, and it will keep reporting for weeks while doing nothing useful.
You found out because something happened to look. On a local disk nothing would have looked, and the first time anybody checked would have been the day the files were needed.
You only know a backup works once it has been read back. That goes for Time Machine on a USB drive, for
rsync on a schedule, and for a folder syncing to the cloud. What helps is a tool that reads
the copy back and tells you when it stops matching.
Copy Shelter does that on a schedule: every piece is checked against the hash it was stored under, a mismatch is reported as a fact, and a piece deleted underneath it is reported as missing. It writes to a drive, a NAS, or a folder in a cloud account you already pay for, and it is free for one location. If you would rather do it yourself, the commands are here.
Usually not. The check that failed is about the backup's internal consistency, and most of the data is normally still readable. Mount the disk in Finder, open Backups.backupdb, and drag out what you need before you do anything else to it.
No. Apple ships no repair tool and there is no supported way to rebuild the index. That is why every answer you will find says to start again, and it is not people giving up early.
It only appears for backups on a network share. Hold Option and click the Time Machine menu bar icon. On a local USB or Thunderbolt disk macOS does not offer it, because it does not run that check on local disks at all.
Only after you have satisfied yourself the hardware is sound. Run diskutil verifyVolume on it and check its SMART status. If the disk is fine, erasing it and starting a new backup is reasonable. If it is not, a new backup on a bad disk fails the same way in a month.
On a network share: keep the NAS and the Mac awake during backups, give the sparse bundle room, and exclude the backup disk from any antivirus scanner. More generally, one backup is one point of failure. A second copy somewhere else is the only thing that turns this from a disaster into an inconvenience.
It can have the same underlying problem, because disks and shares fail the same way for everybody. The difference is when you find out. Verification runs on a schedule, it names the pieces that no longer match, and it keeps the copies it has actually checked separate from the ones it has not.
Related: how to check a backup is actually readable, and the 3-2-1 rule, which is really about this exact morning.