AI File Renaming Statistics 2026

Documents and images make up 94% of everything renamed with AI on Mac. Documents lead slightly, at 49% of files: scans, invoices and downloads with names like scan0001.pdf that say nothing about what is inside.

• DocumentsPDFs, scans, office files, text49%
• Imagesphotos, screenshots, camera RAW45%
• VideoMP4, MOV, MKV and more4%
• Audiorecordings, voice memos, music2%
• Design filesSketch, Figma, Photoshop<1%

8% of batches mix several file types in one go, for example a downloads folder with PDFs, screenshots and a video. Shares are by number of files, Mac only: the Windows app does not report file types.

Most renaming is small: the median batch is 11 files, and 62% of batches contain 20 files or fewer. Large clean-ups are rare but carry most of the volume. Batches of more than 500 files are only 8% of all batches, yet they account for 67% of the files renamed.

With cloud AI, the median is about 2 seconds per file. That is the time a person actually waits, from starting the analysis to having names for the whole batch, including the network. Bigger batches are faster per file because files are analyzed in parallel.

By file type on Mac: images take 1.3 s per file, documents 1.5 s and videos 2.6 s, because a video needs frames, speech or subtitles before it can be named. Free and paid users get the same speed (2.1 vs 1.9 seconds per file). Applying the new names afterwards takes about 40 milliseconds per file.

Local models (Ollama, LM Studio) are not included in these timings: their speed depends on the computer and the model: typically several seconds per file, and more than half a minute at the slow end.

Rarely. More than 4 in 5 active Mac users never edited a single AI-suggested name by hand before applying it. Every batch in Zush is shown as a preview first, and names can be changed with a double-click, so this measures people choosing not to, not people who could not.

Counted for Mac users who renamed at least 10 files between September 18 and November 10, 2026. Someone who dislikes a suggestion can also leave the file out instead of editing it; that case is not part of this figure.

98.9% of files sent to batch renaming were renamed without an error. A file that fails keeps its original name and can be renamed again.

Among batches with a known AI mode, about 1 in 11 is processed by a local model running on the user's own computer instead of a cloud provider, mostly through Ollama and LM Studio.

• Source: anonymous product analytics from the Zush apps for Mac and Windows, plus aggregate usage counters. Only counts, file type groups and durations are used. No filenames or file contents are collected.
• Period: August 10 – November 10, 2026, unless a figure says otherwise. Analysis timings start on October 5, 2026, and edit counts on September 18, 2026, when the app began reporting them.
• Durations and batch sizes are medians, because a few very large batches and long videos would distort an average. Time per file is the wait for the whole batch divided by the number of files.
• File type shares are Mac only. Developer test sessions are excluded.
• This is first-party data published by the maker of Zush, not an independent study. For how Zush evaluates naming quality, see the methodology page; for what the app sends and when, see the privacy policy.

The statistics on this page are free to use in articles, videos and research under CC BY 4.0. Please credit the Zush File Renaming Index and link to this page.

Zush File Renaming Index 2026, https://zushapp.com/ai-file-renaming-statistics (data: August 10 – November 10, 2026).

Zush reads what is inside a file and suggests a name for it, then shows the whole batch as a preview before anything changes. It runs on Mac and Windows, with cloud AI or local models.