The 6 AI engines we track, and why those six
A short note on which answer engines we cover by default, why we picked those specifically, and what we removed and why. Mostly an audience and signal-quality argument, not a popularity contest.
Promlo tracks six AI engines on every plan, including Free: ChatGPT, Claude, Gemini, Perplexity, Grok, and Llama. People ask why those six and not others, so this is the short version.
The selection rule is "engines our customers' buyers actually use"
We track answer engines that meaningfully influence purchase decisions among the customers we serve — SaaS teams selling to international (mostly English-speaking, with strong HK / TW / SEA secondary) buyers. The qualifying bar is buyer share, not press coverage.
ChatGPT is the default. Among everything we measure, this one moves the needle most often. If a buyer is going to ask AI a category question, the question is more likely to land here than anywhere else.
Gemini rides Google's distribution. Logged-in Google users get AI Overviews on a growing slice of informational queries. Android-default Gemini installation pushes mobile coverage up further.
Perplexity is the smallest of the four "headline" engines by raw users but disproportionately influential among power users — engineering managers, founders, VCs. Its citation behaviour is also the most measurable, which makes it punch above its weight in our analytics.
Claude is the rising surface in B2B. Less consumer reach than ChatGPT, but heavy use among technical buyers, and the engine improves quickly between training cuts.
Grok is here because xAI's web-search loop got serious enough in early 2026 that ignoring it would mean missing real signal, especially for product-on-Twitter audiences.
Llama is here as the open-weights surface. Llama itself doesn't have a consumer chat app the way ChatGPT does, but it's served through Perplexity-likes, internal corporate chatbots, and OSS interfaces in ways that compound over time.
What we removed and why
We previously had Doubao, Qwen, and Kimi in the engine list because the original positioning leaned more on a "China + Western" story. We dropped them — not as a quality judgement on those engines, but because the buyer we serve doesn't make purchase decisions on them, and citations from them don't transfer to international answer surfaces. Adding engines our customer's buyers don't use makes the dashboard noisier, not stronger.
Bing Copilot is similarly absent. Bing share in HK / TW is negligible, and even in markets where it's stronger, the answer-quality and citation-graph signal is weaker than Perplexity for the same queries. Adding it would burn a column on the matrix without sharpening anyone's decision.
What this means for prompt design
Six engines × N prompts × four runs a month is the working unit. The matrix isn't six brand-equivalent surfaces — they each behave differently, cite differently, and update at different cadences. Tracking them in parallel and watching where you appear (and where you don't) on the same prompt across all six is the analytics insight that none of the single-engine browser-tab workflows can give you.
If you'd rather build all of this yourself, How to track ChatGPT mentions of your brand walks through the workflow honestly. If you'd rather use a tool that already does it, the Free tier covers all six engines.