Mentioned isn't recommended: how ChatGPT and Gemini pick HR software
- Finding 01Deel was named in 68% of 60 AI answers but made the top three only once.
- Finding 02Rippling was ranked first in 52 of 60 answers, across both assistants.
Your next client will ask ChatGPT, Perplexity or Google before they ever open your website. We engineer how search engines and AI assistants understand your business, so you get found, cited and recommended.
Top picks: HubSpot, Attio … [1]
Top picks: Rippling, Gusto … [1]
Top picks: Freshdesk, Zoho Desk … [1]
Top picks: Linear, ClickUp … [1]
Ranking on Google is still the base. But buyers now get answers without clicking, from engines that decide which brands to name. Each surface reads and rewards something different.
| Surface | What it reads | What gets you in |
|---|---|---|
| Google Search | Your pages, links and structured data | Relevance, authority and a technically healthy site |
| ChatGPT | Live web results plus what it learned in training | Consistent facts about you across many sources |
| Claude | What it learned in training, plus live web search when it needs it | Specific, well-evidenced pages it can trust and cite |
| AI Overviews | Pages that already rank, summarised above the results | Clear, direct answers near the top of the page |
| Gemini | Google's index and Knowledge Graph | A well-defined entity: who you are and what you do |
| Perplexity | Live web search, cited inline | Pages that state things plainly enough to quote |
We work on all six at once, because they draw on the same foundations: a clean site, a clearly defined brand, and content worth citing.
Start with an audit. It shows where you stand and which of the next two steps matters first.
“We don’t know what AI says about us.”
See exactly how AI assistants describe your brand today, and what to fix first.
“AI names our competitors, not us.”
Get named and cited in ChatGPT, Perplexity, Gemini and Google AI Overviews through answer engine optimization.
“Our rankings keep slipping.”
Rankings that hold, built on a site search engines can crawl, parse and trust.
Every engagement starts with a measured baseline, so you can see what changed instead of taking our word for it.
We run a fixed set of buyer prompts across Google and AI engines and record where you appear, how you're described and which sources get cited. {{PH:AUDIT_TURNAROUND}}
Crawl issues, structured data, conflicting brand facts and pages engines can't parse. The unglamorous work that everything else depends on.
Clear answers and original data on the pages that matter, written so engines can quote them and people can act on them.
The same prompts, re-run every month. You see what moved, what didn't, and what we're changing next.
Original studies on how AI assistants choose which brands to name. Method, raw data and limitations included, free to cite.

CobaltKite is run by Tejas D Jaiprakash, with a background in AI and ML engineering. That background shapes the work: we test how engines actually behave, measure it, and publish what we find.
About Tejas D JaiprakashStart with an audit or a short call. You'll get a clear next step within one business day, even if that step is "not yet".