Gokan's Studio

Are SharePoint’s built-in Image Tagging and Taxonomy Tagging services dead?

In modern SharePoint, new AI features like AutoFill columns and the Copilot Knowledge Agent have transformed how content is tagged and classified. AutoFill uses large language models to automatically extract or generate content from files and save it to library columns. You can define a natural-language prompt (e.g., “What are the main categories of this document?”), and SharePoint will fill the answer into a column. Meanwhile, the Knowledge Agent continuously scans your libraries and auto-tags files with AI-generated metadata, including classifications and keywords for images and documents. In practice, you can ask SharePoint (or Copilot) to “tag this image with relevant keywords” or “populate the Category field from this file,” and the AI does it for you. All of this happens natively with your Copilot/Autofill subscription, so for everyday tagging tasks, a separate image- or taxonomy-tagging service is essentially superfluous. In short, built-in AI now covers most scenarios that those old services handled.

• AutoFill Columns: You set up a column prompt (grounded to each file) to classify or extract info from content. This works on any supported file (text, PDF, images, etc.) and can populate text, choice, or even managed metadata columns. In fact, AutoFill now supports managed metadata term sets (up to 100 terms); if a synonym matches, it returns the preferred term. You could ask it to “list terms from our Taxonomy that describe this report” and have them inserted automatically.

• Knowledge Agent: This SharePoint agent can discover and enrich content in the background. It will auto-create and fill columns, including adding tags to images and documents, without prompting the user. Thanks to built-in OCR and vision, it even reads text from images/PDFs and extracts visual features. For example, it can detect objects or scenes in a photo and automatically tag them. Because the Knowledge Agent is included with every Copilot license, all users benefit from its metadata suggestions at no extra cost.

Just bear in mind that if there is no text in your image, Autofill can throw you such errors.

Error in DocumentQuestionAnswering. Exception: OneDrive.Media.Utilities.MeTAExceptions.GeneralMeTAException: Document does not

• Copilot and Chat: Copilot’s AI can classify content on the fly. You can open a file or image in Copilot (or even Teams/Office) and ask it, “What tags apply here?” or “Describe this picture,” then use the answer as metadata. Combined with AutoFill and the Knowledge Agent, Copilot effectively sees your content and its tags together. (Microsoft even notes that Copilot will soon reason over file tags and categories, not just raw text.)

Because of this new AI-powered tagging, a standalone image- or taxonomy-tagging engine is mostly obsolete. The AI you already have in SharePoint does the job. It finds keywords, classifies files, and applies taxonomy terms without any manual tagging or custom models. In practice, you don’t need to call a separate “image tagger” or “term matcher” – just let Copilot/AutoFill handle it.

SharePoint’s image tagging feature uses AI to generate descriptive keywords for pictures. For example, after uploading an image, SharePoint might auto-fill the “Image Tags” column with terms like “indoor”, “group”, or “diagram”. This makes images easy to search and filter. AutoFill can achieve a similar result by using an image-analysis prompt: you could ask it to “extract any visible objects or text from this photo” and save the output in a column. In both cases, AI is handling the tagging.

Red= OCR + Image Tagger

Green = Autofill (only one had OCR, so autofill without text isn’t that great for explaining images, huh 😊)

Let’s now add a “Image Tagger” Autofill Column and see what Autofill will extract.

As you can see, the image tags of Autofill are way better than the legacy 😊

Legacy Tagging Services | The Last Stand

That said, Microsoft’s classic Syntex tagging services still exist – but mainly for bulk or specialized scenarios. Taxonomy Tagging will auto-apply terms from your Term Store to files in libraries. It’s a simple, no-training approach: you enable it on a site, and newly uploaded Word, PDF, or PowerPoint files get tagged with matching term-store entries. This is handy for letting your existing taxonomy enrich content without writing prompts. Similarly, Image Tagging uses Azure Vision AI to label images with keywords. It automatically writes those labels into an “Image Tags” metadata column in the library. Both features require linking an Azure subscription (Pay-as-You-Go) and turning on the service in the admin.

However, these legacy services have limitations. Taxonomy tagging only works on certain file types (native .docx, .pdf, .pptx) and can take 20–24 hours to update tags, since it processes files in the background. Image tagging only processes new pictures and may take minutes or hours to finish. You also get a static managed column that you can edit, and you pay per document/image processed. By contrast, the AutoFill approach is immediate and interactive (and included in your seat license).

In summary, the old Syntex taggers are in their last stand. They still work for large-scale tagging jobs, but for everyday library use, most organizations will rely on the built-in AI. The Knowledge Agent can even auto-fill date, amount, or other fields from scanned documents (using OCR), a task once reserved for the separate OCR/document model service. In effect, almost everything those legacy AI services did can now be done by Copilot’s agents.

Comparing Approaches

Feature	Copilot/AutoFill (LLMs & Agents)	Syntex Tagging Services (Legacy)
Primary Use Case	Real-time or background AI tagging: classify images and docs on the fly via prompts or agents. Enrich metadata instantly for search/grouping.	Batch or library-wide tagging: apply term-store metadata or image labels to files. Suitable for pre-configured bulk tagging jobs.
Underlying Tech	Large Language Models (GPT-4, etc.) for text/classification, plus Azure Computer Vision for images (via Knowledge Agent)	Dedicated document processing: Taxonomy tagging scans text for term matches, Image tagging uses Azure Vision–powered AI.
Cost/Billing	Included with Microsoft 365 Copilot license (seat-based; no extra per-use fees).	Pay-as-you-go Syntex billing: charged per document or image processed (requires an Azure PAYG subscription).
User Experience	Set up via natural-language column prompts or the “Organize” agent UI. Returns metadata instantly (seconds) in-line.	Enabled in library settings. AI tags are added asynchronously (minutes to hours). Users see results in dedicated “Image Tags” or a configured taxonomy column.
Supported Formats	Works on common file types (Office docs, PDFs, images). Autofill even supports Managed Metadata columns.	Image Tagging supports popular image formats; Taxonomy Tagging supports native Word/PDF/PPT. Scanned or encrypted files may be skipped.
When to Use	Ideal for most tagging: quick setup, flexible prompts, and instant updates. Perfect for Copilot-enabled orgs.	Use if you need to bulk-tag thousands of files against an existing taxonomy, or to tag images at scale retroactively.

Conclusion: Dead… or Just Sleeping?

So, are image-tagging and taxonomy-tagging “dead” in SharePoint AI? Mostly yes. For the vast majority of scenarios, Copilot and the Knowledge Agent have rendered the old services redundant. You can automatically classify content and extract keywords without any manual effort, using the AI you already have. Even Microsoft notes that the Knowledge Agent’s AutoFill has effectively subsumed these “traditional” AI tag features. What remains of the old Syntex taggers are niche cases: full-library batch tagging or preserving legacy workflows.

In practice, organizations with Copilot seats now get automated image and taxonomy tagging for free (I know Microsoft doesn’t like it calling it this way – so let’s say it’s included).

This shift led me to think that Microsoft may eventually merge or retire the standalone tagger features altogether. Why maintain a separate “image tagger” that charges per image, when Copilot’s LLM can do it instantly at no extra cost? In any case, SharePoint users should invest in the Copilot/Knowledge Agent tools for tagging needs. The old AI services aren’t being updated, and their core capabilities are already built into the Copilot experience.

Gokan Ozcifci

In short, for everyday image and metadata tagging in SharePoint, the new AI does it all. The legacy services can quietly fade away — unless you really need that one last round of batch processing.

Hope this helps!

Renewed Revolution!