The rank tracking API
for AI search.
How often ChatGPT, Gemini, Claude, Perplexity and the rest of the eight engines Honeyb tracks recommend you, who they cite, and what to do about it today. Clean JSON, evidence attached.
ChatGPT
Gemini
ClaudeWhy this API exists
Every data API ends at a chart. This one ends at an action. The same engine that tells Honeyb customers what to do each morning, exposed over HTTP.
7
endpoints, from raw mentions to a ranked next action
Daily
scheduled scans; every aggregate ships its sample count
1 GET
between your product and your customers' AI visibility
The headline endpoint
Recommendations, over HTTP.
Honeyb weighs everything it measured today, ranks what would move your AI visibility most, and hands you the action with its reasoning and evidence. The platform shows it on a dashboard; the API hands it to whatever you are building.
- Ranked, not raw. One next-best action per day, ordered by expected impact, instead of forty charts to interpret.
- Reasoned, not oracular. Every recommendation carries its why: the prompt, the gap, the sources driving the current answer.
- Executable. When the action is content, a draft from the content engine can come with it.
Endpoints
Seven ways to ask one question: does AI recommend you?
From the answer a buyer sees, down to whether the crawler behind it could reach your site, up to what to do about it.
Today's ranked next-best action with the evidence attached. The endpoint the rest of the category does not have.
Recommend rate, average position and presence per engine, from repeated daily samples rather than one-off checks.
Every answer that named your brand: the engine, the prompt, the full text and how you were described.
The sources shaping your category's answers: which domains engines cite, how often, and whether you are in them.
Your slice of the answer against named competitors, tracked over time per engine and per prompt group.
How the engines frame you: positive, cautious or negative, with the answer text as evidence.
Whether GPTBot, ClaudeBot, PerplexityBot and friends can physically reach your site, re-tested daily.
The measurement layer
Numbers you can put in front of a client.
Recommend rate is the metric this category was missing: of the times buyers asked, how often were you the answer? Because AI engines answer differently run to run, every figure is computed across repeated samples, and the sample count is part of the payload. No mystery scores, no unexplained indexes.
Build with it
Wherever you are building, the recommendation goes in.
Inside your product
Building a marketing platform, a CMS, an agency portal? Serve AI visibility and next-step recommendations to your users as a native feature, without building a five-engine sampling pipeline.
Inside your agents
Your automations and AI agents can pull today's recommendation, act on it, and re-check the recommend rate after. A closed loop with no human copy-pasting between dashboards.
Inside your reports
White-label the whole story for clients: here is where you stand, here is why, here is what we are doing next. The reason column comes straight off the endpoint.
Inside your alerts
Recommend rate dropped, a competitor took your slot, a crawler got blocked by a CDN update. Route it to Slack or a ticket the day it happens, not at the monthly review.
Why it holds up
Built for a surface where the same question gets different answers.
Sampled, not spot-checked
AI answers change between identical runs, so single checks are noise. Every number is computed from repeated scheduled samples, and the sample count ships with the response.
Evidence included
Aggregates come with the receipts: the underlying answers, the prompts, the cited sources. Your integration can show the why, not just the score.
Decisions, not just data
The recommendations endpoint returns what to do next, ranked, with the reasoning attached. Data APIs end at the chart; this one ends at the action.
FAQ
Frequently asked questions.
Free to start
Get recommended by AI search models.
Run a free AI search visibility check, then let the Honeyb agent do the work that gets you into the answers.
ChatGPTClaudeGemini