Filter Search Results by Publication Date with Seltz

Date filtering is live: control how fresh your search results are

When you're building an AI agent that reasons over the news, freshness determines whether the answer is useful or already wrong.

A customer building real-time AI workflows on top of Seltz came to us with a specific need. They wanted to constrain results to a narrow time window: sometimes the last 24 hours, sometimes a specific day in the past. The query worked and the results were good, but they were sifting through too much noise to surface the moment that mattered.

We shipped date filtering.

How it works

Two new optional parameters on the search endpoint: start_published_date and end_published_date. Both accept ISO 8601 timestamps. Pass either, both, or neither.

bash

curl -X POST https://api.seltz.ai/search \
  -H "Authorization: Bearer $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "xAI Anthropic compute deal",
    "max_results": 10,
    "filter": {
      "start_published_date": "2026-05-07T00:00:00Z",
      "end_published_date": "2026-05-07T23:59:59Z"
    }
  }'

bash

curl -X POST https://api.seltz.ai/search \
  -H "Authorization: Bearer $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "xAI Anthropic compute deal",
    "max_results": 10,
    "filter": {
      "start_published_date": "2026-05-07T00:00:00Z",
      "end_published_date": "2026-05-07T23:59:59Z"
    }
  }'

bash

curl -X POST https://api.seltz.ai/search \
  -H "Authorization: Bearer $SELTZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "xAI Anthropic compute deal",
    "max_results": 10,
    "filter": {
      "start_published_date": "2026-05-07T00:00:00Z",
      "end_published_date": "2026-05-07T23:59:59Z"
    }
  }'

The response returns documents within the window, with published_date as a field on every result regardless of whether you're filtering:

json

{
  "documents": [
    {
      "url": "https://example.com/article",
      "content": "...",
      "published_date": "2026-05-07T14:32:00Z"
    }
  ]
}

json

{
  "documents": [
    {
      "url": "https://example.com/article",
      "content": "...",
      "published_date": "2026-05-07T14:32:00Z"
    }
  ]
}

json

{
  "documents": [
    {
      "url": "https://example.com/article",
      "content": "...",
      "published_date": "2026-05-07T14:32:00Z"
    }
  ]
}

The field is exposed on every document, not just filtered ones. Even if you're not filtering, you can read the publication date off every result and make decisions downstream: sort by recency, score relevance against time, or use it however your agent needs.

Publication date, not crawl date

When we scoped this, the customer told us publication date was the priority. Crawl date was a distant second: useful when publication wasn't available, but not the primary signal.

Publication date is what the world cares about. Crawl date is an artifact of our infrastructure. We use publication date when it's exposed by the source and fall back to crawl date when it isn't, but the field you're filtering against represents when the article actually went live.

Try it in the playground

The Search Playground in your console now includes From Date and To Date pickers. Drop in a query, set a window, run it. No code required to test.

What's next

Date filtering is the first of a set. Domain filtering, the ability to include or exclude specific publishers, is in flight and ships next. More filter types are queued behind it.

If there's a filter you want, tell us. We prioritize what we hear from customers.

Try date filtering in the playground