Which LLM trackers let me track custom prompts? What we look for on client programs
Which LLM trackers let you track custom prompts? What we check before choosing one for client work: caps, personas, countries, and the research behind each prompt.
Nearly every LLM tracker lets you add a prompt. That is not the question we ask when we choose one for a client. We ask who writes the prompt set, how many we can run, how we can slice them, and how we know which prompts deserve to stay.
We run client programs on Promptwatch, mostly because of how it handles that last point. Here is the checklist we use, and how a custom prompt set gets built in practice.
Where custom prompts come from
The worst prompt sets are the ones a tool generates on its own, because they reflect the vendor's guess about your market. The best ones come from sources the client already owns:
- sales call notes and the questions prospects ask in the first meeting
- support tickets and chat transcripts
- Search Console queries that are phrased as questions
- People Also Ask boxes for the client's key topics
- competitor comparison pages and their FAQ sections
Promptwatch's Prompt Explorer pulls suggestions from Search Console, ranking keywords, People Also Ask or any page we point it at, which saves the first manual pass. We still edit every prompt by hand, because suggestions are starting material, not a finished panel.
How we structure the set
A flat list of 200 prompts is hard to read. We organize by:
- Topic. The subject area, so we can report visibility per topic.
- Prompt type. Organic discovery, brand, and competitor comparison prompts behave differently and should be reported separately.
- Intent. Informational, commercial or transactional, because the answers cite different sources.
- Persona. A CFO and a junior marketer phrase the same need differently. Promptwatch supports personas, including persona stacking, so one prompt can run as several buyer types.
- Country and language. Country-level tracking is on every paid Promptwatch plan. State and city tracking starts on Business and is included on agency plans, which matters for local clients.
Tags and topics are cheap to set up at the start and expensive to retrofit later.
Volumes and fan-outs: which prompts are worth keeping
Not every prompt deserves a slot, especially when a plan has a cap. Promptwatch's Prompt Volumes shows search volumes, difficulty scores and citation rates across AI search, Google and Bing, which helps us rank candidates. We also look at query fan-outs, the searches a model runs behind a prompt, because they show which sub-questions the answer depends on. A prompt whose fan-outs hit pages we cannot compete with is a lower priority than one that exposes a gap we can fill.
Caps that matter
Caps are where the plans differ most, so we scope this early:
- Explore (free): 10 prompts, ChatGPT only. Good for a first look.
- Essential ($95 a month): 50 prompts.
- Professional ($245): 150 prompts.
- Business ($579): 350 prompts.
- Agency plans (Kick-off $199, Growth $399, Scale $799): unlimited prompts and unlimited projects, with response limits instead.
For agency work, the unlimited prompts on agency plans are the main reason the economics work. We would otherwise be rationing prompts per client. For a single brand with a large product catalog, 350 on Business can also be tight, and we plan the split between core prompts and rotating exploratory ones.
Other trackers: what we have noticed
Other trackers have their own limits, and we are cautious about generalizing. Some set a fixed number of prompts on entry plans, such as 125 prompts on one vendor's $250 plan. Some meter by credits, where the number of prompts you can really run depends on how many engines you track. Some offer a free trial with a recommended prompt set only, which is not a custom set at all. Some include a small daily quota of custom prompts with a broader SEO subscription. Before committing a client, we read the plan page for the real limits and test with their own questions, not the sample ones.
What a prompt tracker cannot do
It cannot tell you what real users ask. It tells you what your chosen prompts return. That is why the quality of the prompt set matters more than the number of prompts. A set of 40 well-chosen questions beats 400 generic ones, and the results are still samples: answers change between runs, so we read trends over weeks.
Our advice
Write the prompts from client sources, group them by topic, type, intent and persona, keep the set stable for a month, and add exploratory prompts in a separate group. Choose a tool whose caps fit the structure you want. If you are not sure, Promptwatch's free Explore plan lets you set up 10 prompts and see how organization and reporting work before you scale.