The Complete AI Keywording Workflow for Stock Photos
Last updated: July 2026 · 7 min read
Keywording is the least glamorous part of stock photography — and the one that decides whether your images ever get seen. A photo with weak metadata is invisible, no matter how good it is. Done by hand, a 200-image shoot takes a full day to tag. Done with a modern AI workflow, it takes minutes. Here is the exact process, from upload to platform-ready export.
Why manual keywording doesn't scale
Most contributors spend 5–20 minutes per image writing a title, a description and 30–50 keywords. Across a batch that adds up fast, and quality drops as fatigue sets in — the 150th image of the day gets half the keywords of the first. Worse, manual keywording is inconsistent: the same subject gets tagged differently across a portfolio, which fragments your search visibility.
Spreadsheet-based batch editing helps with consistency but not with quality — you end up copy-pasting the same generic keyword block onto very different images, which agencies increasingly flag as spam.
The 5-step AI workflow
1. Batch upload your shoot
Drag the whole batch in at once — JPEGs straight from export. Videos work too. There's no need to pre-sort; the AI analyzes each file individually, so mixed-subject batches are fine.
2. Generate metadata for the whole batch
One click generates a title, description and up to 50 keywords per image, ordered from most to least specific. A 120-image batch typically finishes in under five minutes.
3. Review the outliers, not everything
Don't proofread all 120 results — that defeats the purpose. Scan for images the AI may have misread: abstract shots, heavy post-processing, niche technical subjects. Fix those few by hand.
4. Check for rejection triggers
Before export, scan metadata for trademarks, brand names and terms that require releases. Catching a stray "iPhone" or "Nike" in your keywords now saves a rejection later.
5. Export in your platform's format
Export a CSV for Shutterstock, Dreamstime or Getty / iStock, a .txt for Adobe Bridge, or XMP sidecar files for Lightroom. Upload the images plus the metadata file — no retyping anywhere.
Match the export to the platform
Every agency ingests metadata differently, and using the wrong format is the most common reason batch uploads silently lose their keywords:
- →Adobe Stock — upload via Adobe Bridge with a .txt metadata file, or paste per image in the contributor portal
- →Shutterstock — CSV with filename, description and keywords columns
- →Dreamstime — CSV with its own column layout (title and description are separate fields)
- →Getty / iStock — CSV through ESP or DeepMeta
- →Lightroom / Bridge workflows — XMP sidecar files, so metadata travels embedded with the image
Quality control: what to actually check
AI gets the visual content right far more often than a tired human does, but it can't know your intent. Spend your review time on three things: is the first keyword the term a buyer would actually type; does the title describe the use case, not just the objects; and are there any terms in the list that aren't visible in the frame. Those three checks catch nearly every problem worth fixing.
How AI Keyword Genius fits this workflow
The whole pipeline above runs in one place: batch upload, per-image titles, descriptions and up to 50 ordered keywords, automatic rejection-trigger scanning, and one-click exports for Shutterstock, Dreamstime, Getty / iStock, Adobe Bridge and XMP. A batch of 120 images is typically tagged and export-ready in under five minutes.
Summary
- ✅Manual keywording doesn't scale — quality drops with every image in the batch
- ✅Upload the whole shoot at once and generate metadata in one pass
- ✅Review outliers and rejection triggers, not every single result
- ✅Always export in the exact format your platform ingests
- ✅Budget minutes per batch, not minutes per image