Is this an endorsement or ranking of product quality?

No. We measure MENTIONS, how often engines name Portkey, not whether it is a good product. This page is measurement data, not a review.

Who publishes this and why should I trust the numbers?

CrediGeo, an AI-visibility measurement service. You don’t have to trust us. Every category page links its raw runs (prompts, answers and citations) for download, and the method is public. Ask an engine one of those prompts today and it will name a different set. We publish the archive rather than promise a repeat, because that drift is the thing being measured.

I work at Portkey, can we see the detail behind this?

Yes, request the free breakdown. Which exact prompts named you, which didn’t, per engine, with the sources each answer cited. Free, no strings.

AI Visibility Index · vendor profile · August 2026

Portkey: how often the AI engines actually name it

As of August 21, 2026, Portkey was named in 48 of 97 measured AI answers across 2 categories. Strongest in LLM gateway (AI model routing and proxy) (46 mentions). Engines are reported separately below; they rarely agree.

Per category, per engine

Cells read “named / answers shown.” Each category links its full leaderboard and raw data:

Category	ChatGPT	AI Overviews	AI Mode	Gemini	Stability
LLM gateway (AI model routing and proxy)	20/20	3/6	11/12	12/12	1 of 16, chance predicts 3.6
LLM observability (tracing, evaluation and monitoring for LLM/agent apps)	2/20	0/3	0/12	0/12	1 of 16, chance predicts 1.9

Source note: across these measurements, the engines cited Portkey’s own domain 31× while building answers, labeled “vendor-owned” in the category source tables, because self-published pages are a different kind of evidence than independent ones.

measured 2026-08-21 · aggregated from 2 categories · re-measured monthly

Read this honestly

These are dated snapshots of a moving target: AI answers change run to run and month to month, and mention counts are not quality judgments. Portkey did not pay to appear here and cannot pay to change these numbers, the data re-measures on schedule either way.

Work at Portkey? Get the full breakdown behind these numbers. Which exact prompts name you, which don’t, per engine, with cited sources. Free, in your inbox within 2 business days.

Request your breakdown

Questions about Portkey's AI visibility

Does AI recommend Portkey?: In our August 21, 2026 measurement, Portkey was named in 48 of 97 AI answers across 2 measured categories, strongest in LLM gateway (AI model routing and proxy) (46 mentions). AI answers vary run to run and engine to engine; the tables here show the full split.
Is this an endorsement or ranking of product quality?: No. We measure MENTIONS, how often engines name Portkey, not whether it is a good product. This page is measurement data, not a review.
Who publishes this and why should I trust the numbers?: CrediGeo, an AI-visibility measurement service. You don’t have to trust us. Every category page links its raw runs (prompts, answers and citations) for download, and the method is public. Ask an engine one of those prompts today and it will name a different set. We publish the archive rather than promise a repeat, because that drift is the thing being measured.
I work at Portkey, can we see the detail behind this?: Yes, request the free breakdown. Which exact prompts named you, which didn’t, per engine, with the sources each answer cited. Free, no strings.

Get your free breakdown

A run is one question, asked once, in a clean session. We ask your buyers’ questions on ChatGPT and on Google’s AI Mode, more than once on each, then send you who got named in every answer with the exact prompts.

What arrives, exactly

Engines: ChatGPT and Google AI Mode
Runs: 3 ChatGPT and 2 AI Mode per question, each a fresh session
You get: Every answer, who was named in it, and the exact prompts
Arrives: Within 2 business days, from a person
Cost: None, and no call required

Prefer to talk first? Book a walkthrough (opens in a new tab). Your category on screen, no deck. Or read the real sample report.