Are you still using the Meta field on Knowledge Articles?

[RogueFader]

Hi all

We're using ServiceNow Knowledge with Now Assist / AI Search, and historically we've populated the Meta field with keywords to improve findability.

Since rolling out AI powered search, it seems the platform is finding articles from the article content itself rather than relying on metadata.

I'm interested to know if people are still using the Meta field? and If so, what are you storing in that field? is this a move away from keyword tagging?

It was discussed about how valid it is these days and do we even need it since we have GenAI, however in the back of my mind, I'm still reluctant and thought we could still retain it for taxonomy, reporting and governance (e.g. Service, Service Offering, Product, Audience, Knowledge Type) rather than traditional keyword search

I'd be interested to hear how people are approaching metadata in a AI ServiceNow Knowledge environment and whether you've found it still adds value.

Thank you.

[nilskrause]

think your instinct is basically right, but I would make an important distinction:

AI Search reduces the need for metadata as a keyword-search mechanism. It does not make metadata obsolete as a knowledge-management mechanism.

I checked the current ServiceNow documentation and recent ServiceNow Community discussions, and the picture in 2026 is actually quite nuanced.

1. The Meta field is not obsolete

Current ServiceNow documentation says that AI Search indexes text/string content and metadata from indexed sources.

And importantly, a recent ServiceNow Community discussion confirms that the Meta field is now included in AI Search indexing by default; this wasn't always the case. The recommendation there is to check your particular indexed-source field configuration to verify that Meta hasn't explicitly been excluded.

So I would not remove your Meta field simply because you're using AI Search/Now Assist.

But I would change how you use it.

2. I would stop thinking of Meta as "keyword stuffing"

This is the biggest change I would make.

Historically:

Article: VPN connection problems
Meta: VPN, VPN problem, VPN connection, remote access, tunnel, Cisco AnyConnect, AnyConnect problem, VPN failure...

The purpose was essentially:

"If the user types one of these words, make this article appear."

That made sense with traditional search.

AI Search is increasingly capable of understanding meaning and context, rather than simply looking for exact keyword matches. ServiceNow's current semantic vector search documentation explicitly describes this difference: semantic search looks at the meaning/context of the query rather than relying solely on keyword matching.

So you don't want to maintain 30 synonyms for every article forever.

3. But Meta still has a very good use case

I would use Meta for information that is relevant to retrieval but doesn't naturally belong in the article itself.

For example:

Article

Title:

How to connect to the corporate VPN

Content:
Explains how to install and configure the VPN client.

Meta

I'd consider things such as:

AnyConnect, VPN client, remote access, corporate network, offsite access, virtual private network, VPN connection

But only where those terms are genuinely useful alternative ways of describing the subject and aren't already adequately represented in the title/content.

A current Community discussion makes exactly this point: avoid duplicating words already present in the article; use Meta for terms, keywords and phrases that aren't already in the article.

4. Your idea about taxonomy is even more interesting

This is where I think your thinking is stronger than the traditional "Meta = keywords" model.

I would separate three concepts:

Purpose Example Where I'd put it
Content How to configure VPN Article body
Retrieval enrichment AnyConnect, remote access, offsite VPN Meta / tags
Knowledge governance ITSM, VPN Service, Employee, How-to Structured metadata

And I would not put all three into the Meta field.

For example, if you have:

Service: VPN
Service Offering: Remote Access
Product: Cisco AnyConnect
Audience: Employees
Knowledge Type: How-to
Owner: EUC
Lifecycle: Operational

those are much more valuable as structured fields / controlled taxonomy than as:

VPN, Remote Access, Cisco, AnyConnect, Employee, How-to, EUC...

inside Meta.

That's because structured metadata can support filtering, reporting, governance, ownership, lifecycle management and analytics, whereas Meta is fundamentally a searchable text field.

5. So I'd use a two-layer model

For a modern ServiceNow Knowledge environment, I would recommend:

Layer 1 — AI-readable knowledge

Focus heavily on:

excellent title
problem statement
natural language
clear symptoms
resolution
consistent article structure
meaningful terminology

This is increasingly important because AI Search can understand the content itself. ServiceNow's current documentation also describes machine-learning relevance tuning that uses aggregated user interaction data and testing to improve result ranking.

Layer 2 — structured knowledge metadata

Maintain:

Service
Service Offering
Product/Application
Audience
Knowledge Type
Category
Ownership Group
Language
Lifecycle/Validity

This becomes your Knowledge Management taxonomy, not merely a search hack.

6. And I'd keep a small Meta field

I wouldn't abolish it.

I'd establish a simple governance rule:

Meta contains only additional terms that help users find the article when those terms are not naturally represented in the article.

For example:

Article title:

Request access to SAP

Content:
Explains the access-request process.

Meta:

SAP access, SAP authorization, SAP permissions, SAP account, SAP rights

That's reasonable.

But I wouldn't have:

SAP, access, request, authorization, permissions, account, rights, application, IT, user, employee, system...

That's essentially keyword stuffing and adds little value.

7. One thing I'd test before changing your KB governance

This is important for your environment.

Don't assume:

"AI Search finds it from the article, therefore Meta isn't doing anything."

You need to look at your AI Search indexed source configuration and relevance behaviour.

ServiceNow's current documentation confirms that indexed sources determine which text/string fields and metadata are indexed.

I'd therefore take 20–30 representative articles and test three variants:

A. Article content only
B. Article + Meta
C. Article + structured taxonomy + Meta

Then test real user queries such as:

"I can't access SAP"

"How do I get SAP permissions?"

"SAP authorization"

"I need access to SAP"

"new employee needs SAP"

Measure:

Does the correct article appear?
Position in results?
False positives?
Search without exact terminology?
Does Meta actually improve ranking?
Does structured filtering help?

That gives you evidence rather than an opinion about whether Meta is still useful.

My recommendation for your BLG Knowledge environment

Given the Knowledge structure you've been working on, I would absolutely retain metadata.

But I'd redefine its purpose:

Old model

Meta = improve keyword search

⬇️

New model

Knowledge content = primary source for AI retrieval

Meta = retrieval enrichment

Structured taxonomy = governance + reporting + filtering

AI Search = semantic/relevance engine

Now Assist = understands/summarizes/uses the retrieved knowledge

That's a much more future-proof architecture.

And there's another reason I like your approach: ServiceNow's newer semantic search explicitly reduces dependence on manually maintained synonym dictionaries because it can identify similar meanings automatically.

So I wouldn't throw away the metadata you've historically built. I'd clean it up and repurpose it.

In fact, for your KB MVP, I would define a "Knowledge Metadata Standard" with perhaps 8–10 controlled fields, plus a very specific rule for what is allowed in Meta. That would give you a clean governance model rather than every KB author inventing their own keywords.

[drosenst2]

I think meta is critical and we are looking to make it a mandatory field, but it has to be used correctly. As mentioned, we use it for retrieval enrichment - for instance we have terms/spellings that are used in the US, but certain articles pertain to Canada/UK/India audience as well. They spell the same word differently paycheck/paycheque, center/centre, behavior/behaviour, etc. Once I added that variation of spelling to the meta, the Canada/UK/India audience was able to find their information easier. Also, we've used search data from end users to update the articles based on the terms/phrases that people are actually using that don't reside in the article either. I also think of meta as how is the end user actually looking for these articles and then account for it there. We also changed meta to a minor change when updating so that you don't have to check out the article in order to change it. Meta is a live and constantly updating field. 

Regarding the front end taxonomy, I agree about filtering, reporting and governance. It can be helpful for new employees or accidental learning (which is always fun). Some people still do browse too especially with certain generations. 

Just my two cents 🙂

[Eoghan Sinnott]

Agree with Meta being critical. With employees across North America, Ireland and Asia we also have the terms/spelling challenge and have used it for this. We also use it for applications/services that may have been rebranded or had a name change, and people often still refer to them by the legacy name.

A good tip is to also use the Synonym dictionary for this if you are using AI Search. Just saves some effort in having to add the same meta data to all relevant articles. You set the synonym once on the search profile, can add multiple entries and results will come back once somebody uses one of the words that has been added!

[brianlan25]

I guess ServiceNow even changed there mind about the meta field. I was questioned once about it before because there was a message that appeared on the screen saying not to use meta with AI Search. But that no longer displays when creating a knowledge article.

However DOCS says otherwise.
Tags for optimizing search from external search engines. Separate multiple entries by commas. You may need to configure the form to add this field. For more information, see Configuring the form layout.

Note: 

If AI Search is enabled for the Knowledge Management Service Portal, the tags you add to the Meta field aren't indexed. To fine-tune AI Search results, your administrator can configure result improvement rules.