We built a feature even though it technically hurts our bottom line. Here's why.

Most AI search vendors (including us) use prompt-based pricing. This creates a financial incentive for vendors to push you toward over-tracking.

More prompts means more revenue for them, whether or not those prompts actually provide anything useful for you.

We’d rather help you measure what matters than run up a bigger monitoring bill. Topic Prompt Optimizations is how.

We designed this feature to help you find your prompt-tracking Goldilocks zone—the ideal number and mix of prompts for each topic you care about.

The goal? Maximum coverage without waste, noise, and redundancy.

Scrunch customers can use it to:

• Identify which prompts contribute unique value and which are unnecessary.
• Develop an optimal set of prompts to monitor for all key topics.
• Customize prompt strategy based on what matters to you.

Here’s how it works.

Keep your valuable prompts, cut the rest

Prompt bloat isn’t just a billing issue. It's also a signal quality issue.

When you track redundant prompts, you get noise, not more insight. We built Topic Prompt Optimizations to identify which prompts are actually pulling their weight.

Not so many that you're paying for overlap. Not so few that you're missing coverage. Just the right amount, all doing distinct work, with nothing wasted.

Get started by navigating to the Insights tab in Scrunch and selecting Topic Prompt Optimizations.

Choose a topic (we populate options based on the topics you add to your AI Context tab) and which AI platforms to analyze (we evaluate prompt performance from the past 14 days across all active models by default).

Then click the “Run optimization” button and we’ll generate your results.

Here’s what you get back:

• All the prompts you’re tracking that ladder up to that topic
• Which prompts we recommend you keep
• Which prompts we recommend you cut

Our out-of-the-box recommendations balance two things: coverage (the percentage of unique URLs cited by at least one prompt) and resilience (the percentage of URLs cited by two or more prompts).

Both are important. Coverage helps you observe the full space of URLs competing for a prompt and resilience helps you understand how some URLs tend to cover more ground in a topic than others.

If new prompts introduce new URLs, you’re expanding your visibility. If they return the same URLs, your prompts are overlapping and becoming redundant.

We also factor in anticipated changes to brand presence, answer position, and sentiment before recommending anything for removal.

You can dig into the results to see which personas, custom tags, and funnel stages each prompt connects to.

If every prompt you're tracking adds value—or if removing prompts would hurt your coverage—you'll get a message like, "Your prompt setup is already efficient" or "Pruning not recommended."

That's a good outcome. It means your setup is working.

But if it's not, you can archive suggested prompts right from your results.

Want to look closer before making any changes? Click into any individual prompt from the results view.

You can also export everything as a CSV, including which prompts to keep versus cut, the closest replacement prompts for any recommended removals, and the configuration used to generate your results.

Dig into the data behind recommendations

Topic Prompt Optimizations is built for everyone, from the "just tell me what to cut" crowd to the data nerds among us who want to see exactly how every decision gets made.

Recommendations are based on a Pareto frontier (a fancy way of saying “the set of optimal prompt combinations where you can’t improve coverage without hurting resilience and vice versa”).

Our default, pre-selected configuration is what we’ve observed work best for bloated topics.

In our analysis, any topic below 95% coverage results in core metrics (e.g., brand presence rate, citation share, sentiment, etc.) fluctuating meaningfully.

So we set your coverage floor—the minimum percentage of unique URLs that must remain cited by at least one prompt after pruning—at 95%.

Meanwhile, we automatically balance coverage and resilience equally.

All of this happens behind the scenes, but we make it very easy to pop the hood and fine-tune your approach.

At the top of the screen you’ll find a dropdown tab for Advanced Configuration.

Click it to see various options for how to optimize your set of prompts: different constraints, a customizable coverage floor, and pre-built pruning strategies.

Select the options you want, click “Apply Changes,” and we’ll update your results.

Meanwhile, at the bottom of the screen you’ll find a dropdown tab for Data Deep Dive.

Open it to see interactive visualizations from our data science team that show exactly how changes to your prompt strategy affect coverage and resilience.

Play around with different configurations until you've found the balance that fits your goals.

You can always click the “Start Over” button at the bottom of the page to begin again.

Start here before you start cutting

Ready to get going? Here are a few tips:

1. Overshoot to start

Start with more prompts than you think you’ll need. The more data you have to work with, the more accurate your results. And our tool makes trimming after the fact easy.

2. Give it time before making cuts

Let your prompts run for at least 2 weeks before using our tool. You need time to tell the story before you can separate valuable prompts from redundant ones.

3. Measure twice, cut once

We recommend using our default pruning strategy (a 50-50 split between coverage and resilience) if you’re unsure where to start.

That said, watch for a flattening coverage curve before pruning. A flattening curve indicates that new prompts are no longer discovering new URLs.

In general, avoid pruning aggressively if every prompt adds unique URLs.

4. Free up prompt budget

Archive prompts that don’t add meaningful new coverage for a topic. It will clean up your setup and also give you room to add new prompts for topics that may not have sufficient coverage.

We’d rather be worth it than just bill for it

AI search vendors have a financial incentive to make your prompt list as long as possible.

We want to fix that.

There’s a fair chance it will shrink your Scrunch bill. But our bet is that you’d rather work with the company that optimizes for your outcomes versus the one that sends you the biggest invoice.

Ready to give it a spin? Reach out to support or your CSM to request access.

Optimize your prompt tracking strategy with Scrunch

Make sure every prompt pulls its weight. Start a 7-day free trial or get in touch to see how Topic Prompt Optimizations can help you right-size your monitoring setup.