---
title: "Kitbase vs LLMrefs: Free Keyword-Style Checks vs a Measured Programme | Kitbase Blog"
description: "Kitbase vs LLMrefs compared: LLMrefs' free multi-model prompt checks against Kitbase's scheduled runs, verified crawler data and conversion analytics."
canonical: https://kitbase.dev/blog/kitbase-vs-llmrefs
---

**LLMrefs is a free way to check whether AI models mention you across a lot of models. Kitbase is a paid system for running that check continuously and explaining the results.** The gap between a spot check and a measurement programme is bigger than it looks, and it comes down to sampling.

We build Kitbase. LLMrefs is free, so for a genuine first look there's no argument against trying it.

## What LLMrefs does well

It generates conversational prompts from keywords and checks visibility across up to eleven models, on a free tier with low-cost paid plans above it. The keyword-derived approach makes it immediately legible to anyone with an SEO background: you know how to think about keywords, so you know how to think about this.

For answering "do AI models know we exist," it works, costs nothing, and takes minutes. That's a real service and plenty of teams need nothing more.

## Where a spot check stops working

Ask ChatGPT the same question twice and the brand list can change. Not occasionally, routinely. The models sample from a distribution, and the surrounding retrieval changes as the web changes.

That has a consequence most people underestimate: a single check is not a measurement. If LLMrefs shows you present in seven of eleven models today, running it again tomorrow could show five or nine without anything having changed on your site or theirs. You cannot tell improvement from noise, which means you cannot tell whether anything you did worked.

Fixing that requires repeated sampling on a schedule, a stored history, and enough runs to establish a baseline. That's what turns a check into a trend, and it's the thing free tools can't afford to give you, because every prompt × engine query is a paid API call to someone.

## What Kitbase adds

**Scheduled, resumable runs.** A run is a batch of prompt × engine queries executed on a schedule. Completed provider calls are recorded and never re-executed, so a deploy or restart mid-run resumes rather than re-paying for finished queries. Runs can be paused and cancelled, and cancelling keeps everything already finished. That matters because the sampling you need for a trustworthy trend is the expensive part.

**Measures defined so they mean something.** Presence rate splits brand mentions from actual domain citations. Share of voice is normalised across every tracked brand with a dense rank over time. Cited sources go to the exact page, classified as yours, a competitor's or third-party and tagged by source type. Every mention carries sentiment, whether you were recommended or merely named, and your position when the answer is a ranked list.

**The layers underneath.** Server-side [AI crawler detection](https://docs.kitbase.dev/crawler-detection) with identity verification and per-path crawl frequency, so you know whether the engines can even read the pages you want cited. Cookieless web analytics on the same pipeline with funnels, journeys and retention, so AI referrals lead somewhere measurable. Backlink opportunities from the domains AI cites where you're absent. A site audit and keyword research.

**The honest limitation:** LLMrefs checks more models than Kitbase's lower plans, and it's free. Kitbase's Starter plan tracks ChatGPT only, at $99 a month.

## Side by side

| | LLMrefs | Kitbase |
|---|---|---|
| Price | Free tier, low-cost paid | From $99/mo, 7-day trial |
| Models checked | Up to 11 | ChatGPT on Starter; up to 10 on Business |
| Scheduled repeated sampling | Limited | Yes, resumable runs |
| Stored trend history | Limited | Yes, per engine and competitor |
| Mentioned vs cited split | Basic | Reported separately |
| Cited-page mapping with source types | No | Yes |
| Sentiment, recommendation, list position | No | Yes, per mention |
| AI crawler detection | No | Yes, with identity verification |
| Web analytics and funnels | No | Yes, cookieless |
| Backlinks / link marketplace | No | Yes |

## Which to use

**Use LLMrefs if** you're finding out whether this channel exists for you, you want to check many models at once for nothing, or AI visibility isn't yet anyone's job.

**Use Kitbase if** someone owns the number, you need to prove a change was real rather than variance, or you want the crawl and conversion data that explains movement.

The honest sequence is LLMrefs first. Establish that AI answers matter for your category, then buy sampling. Paying for a measurement programme before you know the channel affects your pipeline is the wrong order.

## FAQ

**Is LLMrefs really free?**
It has a free tier with paid plans above it. Expect caps on prompts and frequency, which is where the limits of free sampling show up.

**Why does repeated sampling matter?**
Because the same prompt returns different brand lists on different runs. Without repeated sampling you can't distinguish a real change from [normal answer variance](/blog/ai-answers-non-determinism).

**Does LLMrefs track AI crawlers?**
No. It works on the answer layer. Kitbase, Hall, Scrunch AI and Profound report crawler activity.

**Can I use LLMrefs to prove GEO work is paying off?**
Not reliably. Proving impact needs a consistent baseline across many runs, plus referral and conversion data to connect citations to outcomes.

**What's the cheapest way to get real trend data?**
Hall's free tier is the strongest free option that stores history and covers crawlers. Rankscale is the lowest serious paid tier.

---

*Ready to turn spot checks into a measured programme? [Start your free trial](https://app.kitbase.dev/signup/) — 7 days, no credit card required.*