Sonar Pro API misses explicit owner on About page, while browser/Space finds it

Bug Reports
,
1

109

Draft post

Title: Sonar Pro API misses explicit owner on About page, while browser/Space finds it

Hi all,

I’m testing Perplexity for business contact sourcing via API and I’ve hit a consistency issue that looks bigger than prompt tuning.

Scenario

  • Target business: http://www.murwillumbahcombatclub.com.au/

  • Relevant page: /about

  • Use case: identify business owner/operator and return structured JSON

What the browser/Space does

  • In the browser, Perplexity finds the owners correctly.

What the API does

  • Using sonar-pro, the API returns no owner found.

  • The structured output has blank values for:

    • ownerName

    • ownerSource

    • ownerEvidence

  • It sets ownerConfidence to 0.

Why this looks wrong

The About page explicitly says the business is:

“family-owned and operated by locals, Ian McLachlan and wife Gemma”

So this is not a case where the owner is hidden in some obscure external source. The owner signal is right there on the first-party website.

Odd part

The API output includes the About page in pages_searched, but still concludes no owner or proprietor was named on the website.

So it seems the API:

  1. found the page,

  2. looked at the page,

  3. but failed to extract the obvious owner evidence.

What I already tested

  • Same/similar prompt as used in browser

  • sonar-pro

  • higher search context setting

  • confirmed this is not an n8n parsing issue, because the raw API JSON itself has no owner in it

My questions

  1. Is the browser/Space retrieval and extraction path different from the API for this kind of task?

  2. Is there a known issue with API extraction pages?

  3. Is there a recommended API pattern for ownership/contact sourcing where the owner is named on the target website?

  4. Has anyone here seen cases where the API lists the correct page in pages_searched but still misses plainly stated facts on that page?

  5. Is this a Perplexity Bug?

  6. Are there settings I’m missing in the API call?

Practical impact

This matters because for contact sourcing workflows, finding the page is not enough. The API has to reliably extract obvious entity relationships like:

  • owner

  • founder

  • operator

  • director

At the moment, this makes the API less reliable than the browser for structured sourcing.

Recommendation to Perplexity team

A good improvement would be:

  • stronger first-party page extraction for structured tasks,

  • or an API mode that prioritises exact evidence extraction from target pages before broader web synthesis.

Interested to know whether others have hit the same issue.

image
image1920×891 285 KB