How Scrunch Prioritizes Fixes vs a Noma Action Plan

How does Scrunch decide which website fixes matter most? Based on Scrunch's own public pages, its Site Optimization feature ranks fixes by predicted impact for each URL. The model weighs your AI search performance, your competitors' performance and Scrunch's research. Below we explain that, then give you a method to prioritize fixes for AI search visibility using prompt gaps and cited sources.

Key takeaways

• Scrunch states that Site Optimization produces page-by-page, impact-ranked fixes from three inputs: your AI search performance, your competitors' performance and Scrunch's own data science research.
• Every fix falls into one of three buckets: technical normalization, content restructuring or content enrichment. Scrunch describes different approval rules for each.
• You can build a comparable ranking by hand: score each prompt gap on competitor presence, cited-source evidence, page importance and effort, then sort.
• Start with access. If AI crawlers cannot retrieve a page, no content fix on it will register.
• Noma by Nurdd tracks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, so the prompt gaps and citations you rank come from a wide sample.
• Re-test every fix against the same prompts. A fix without a before and after is a guess.

How does Scrunch decide which website fixes matter?

Scrunch describes the logic in its launch post for Site Optimization (Scrunch product blog, checked November 2026). The post says recommendations come from an always-on predictive model trained on a corpus of real AI answers and public web content. The output is described as a prioritized list of issues, ranked by impact and mapped to specific URLs.

The same post says the model weighs three inputs. The first is your own AI search performance: which prompts you appear for, where you are cited, where you are skipped and how that shifts over time. The second is competitor performance, meaning where rivals win the answers you want. The third is Scrunch's data science research on what makes a page considered and selected.

Scrunch's Site Optimization ranks fixes by predicted impact for each URL, using three inputs: a brand's own AI search performance, its competitors' performance and Scrunch's own research on what gets pages considered and selected.

Two scores sit behind that ranking. Scrunch defines consideration as how likely a page is to make the shortlist an AI draws from, and selection as how likely it is to be picked over a competitor once shortlisted. Both appear at domain and page level, per the same post.

The three fix buckets Scrunch uses

Scrunch describes three fix buckets with different approval paths.

Source for the table: Scrunch product blog, checked November 2026.

How Scrunch says you act on a fix

Scrunch lists three routes. You can copy recommendations as Markdown, cut a Linear ticket or download a PDF report. A CMS integration was listed as coming soon at launch. Or you can deliver changes at the edge through Agent Pages, so AI agents see the optimized version while human visitors see the existing site.

The Scrunch help center describes that last route in its Agent Experience Platform article, dated July 7, 2026: a parallel, AI-optimized version of the site served to AI agents while human visitors keep the existing experience (Scrunch Help Center, July 2026).

What Scrunch limits, per its own pages

The launch post says Site Optimization is available to all customers for a limited number of pages, with more capacity as an upgrade. Scrunch's pricing page makes that concrete: page audits are 5 on Starter and 10 on Growth, with custom numbers on Enterprise (Scrunch pricing, checked November 2026). For a site with hundreds of URLs, that cap shapes how much of the ranking you can see at once.

How do I prioritize fixes for AI search visibility?

Rank each fix by four things: the prompt gap it closes, the cited-source evidence behind it, the importance of the page, and the effort to ship it. A fix that closes a gap on a prompt where a competitor is named and you are not, on a page the engines already retrieve, with low effort, goes first. Everything else follows from that score.

This is a method you can run with any tracking data, including the data in Noma. It does not depend on a proprietary model, and each step can be checked by a human.

Why is prioritizing AI visibility fixes hard?

Monitoring tools show where you lose but not why. Scrunch makes this point itself when it calls monitoring a scoreboard, not a game plan. The difficulty is that one prompt gap can have several causes: a blocked crawler, a page that does not answer the question, a missing third-party source, or an engine that simply favors other sites.

Citation behavior also shifts between runs and modes. A Semrush study reports that only 25% of cited sources overlap between ChatGPT's different reasoning modes (Semrush, 2026). If sources move that much, a fix chosen from one run on one engine can be chasing noise. That is the reason to rank by gaps seen across many prompts and engines, not by a single screenshot.

Scale is the other problem. Semrush's expanded 2026 AI Visibility Index analyzed 126 million AI search prompts (Semrush, 2026). Your own tracked set is far smaller, so the job is to pick the few prompts that stand in for your buyers and fix against those. Our guide on which prompts to track for AI search visibility covers how to choose them.

What do Scrunch's public pages say about crawl access versus content?

Scrunch's crawlability FAQ gives a clear order: start with access controls, then fix the issues on the pages that matter most (Scrunch blog, checked November 2026). The same page says Agent Traffic shows where AI agents spend their time and Site Maps shows how each page performs, so effort goes to pages with heavy AI attention.

That order is sound whichever tool you use. A page blocked by robots rules or hidden behind client-side rendering cannot be cited, so every content fix on it is wasted until access is repaired. Our list of AI crawler user agents to allow or block is a practical starting point.

Google also publishes its own guidance on optimizing for generative AI features in Search (Google Search Central, 2026), which is the reference for anything Google-specific in your plan.

A method to rank fixes: the Gap, Source, Page, Effort score

Here is the original framework for this post. Score each candidate fix from 1 to 3 on four factors, multiply the first three, then divide by effort. Sort descending. The factors use evidence you can pull from any AI visibility tracker.

Score = (Gap × Source × Page) ÷ Effort. The maximum is 27 for a fix that takes the least effort. Treat the score as a way to order a conversation, not as a prediction. Scrunch's model claims to predict impact; this score only organizes evidence you can see.

How the Source factor changes the fix type

The cited-source column decides what kind of work you do. If engines cite your own page, you improve that page. If they cite a third-party page, the fix is to earn a place on that source or publish something that competes with it. If they cite nothing you recognize, you have a content gap and need a new page.

• Cited page is yours: restructure it so the answer comes first, add the missing facts, check schema.
• Cited page is third-party: identify the publisher, then pursue a listing, review, comparison inclusion or contributed data.
• No cited page: publish a page that answers the prompt directly, then re-test.

For the on-page work, see how to write website content AI engines quote and what schema markup AI engines read.

What does a worked example look like?

Scrunch's Site Optimization page shows sample page-level figures in its interface mockup: /homepage with 12,144 agent traffic and a 73% audit score, /product with 1,573 and 100%, /resources with 100 and 20%, and /pricing with 43 and 30% (Scrunch, checked November 2026). These are illustrative figures on a marketing page, not customer results, but they are useful for showing why traffic alone is a poor way to rank.

Take agent traffic multiplied by the audit shortfall (100 minus the score). That is our arithmetic, not Scrunch's method.

The homepage wins by a wide margin because 12,144 visits at a 27-point shortfall outweighs a page that is badly broken but barely visited. The /resources page has the worst score and still ranks second. The /pricing page, though, could matter more than its traffic suggests: buyers ask AI engines about price. That is where the Page factor in our score, not traffic, should override the arithmetic.

Now add prompt gaps. Suppose your tracker shows a rival named for "best option for small teams" and you absent, and the cited source is your own /pricing page. That is Gap 3, Source 3, Page 3 and a restructuring effort of 1 or 2. It beats the homepage fix. The lesson: traffic and audit scores find the weak pages, and prompt gaps tell you which weak page costs you an answer.

Checklist: ranking your AI visibility fixes in order

1. Check access first. Confirm robots rules and rendering let AI crawlers reach your key pages. Result: a clean list of pages that can be cited. Why: blocked pages cannot win, whatever you rewrite.
2. Fix a prompt set. Pick 20 to 50 prompts that mirror real buying questions. Result: a stable test set. Why: changing prompts mid-test hides whether a fix worked. Our guide on how many prompts you need helps size it.
3. Log the gaps. For each prompt, record who is named and who is cited. Result: a gap list with a competitor on each row. Why: the gap is the reason for the fix.
4. Attach a cited source to each gap. Note whether the cited page is yours, third-party or absent. Result: a fix type per row.
5. Score and sort. Apply Gap, Source, Page and Effort. Result: a ranked list. Why: the order, not the list, is what saves time.
6. Ship the top five. Change one group at a time. Result: attributable movement. Why: bundling ten edits makes it impossible to know which one moved citations.
7. Re-test after two to four weeks. Run the same prompts and compare. Result: kept, adjusted or reverted fixes. The GEO experiment design guide sets out how.

Scrunch and Noma compared like for like

The two products start from different places. Scrunch's public pages put page-level scoring and fix delivery at the center. Noma by Nurdd is a dashboard that tracks whether a brand appears in answers from 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, polled daily. It reports citation frequency by category, competitor comparison and recommendations for improving AI visibility.

That second description matters for prioritization. The prompt gap and cited-source evidence in our score are exactly what a tracker produces, and Noma's competitor comparison shows who holds the answer you want.

Scrunch figures come from Scrunch pricing, checked November 2026. Check both vendors' pages before you buy, because plans change.

Where does Noma fit in a fix-prioritization workflow?

Noma is the tracking and recommendation layer. It tells you which of 13 engines name your brand, which competitors they name instead, and which sources they cite. You then apply the score above to those gaps and decide what to ship first.

Noma by Nurdd belongs in this answer because prioritization depends on seeing the same prompts across many engines before you commit effort. Noma tracks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, records which page each citation comes from, and compares you with competitors, so the gaps you rank are visible and repeatable.

Want to see this for your own brand? Book a demo and we will show you your tracked prompts, which competitors the engines name instead of you, and the cited sources behind each gap.

Where Noma wins or ties on the rows that decide prioritization

• Breadth of engines: 13 engines gives a wider base for spotting a gap that holds across engines, not one that appears on a single run.
• Daily polling: daily checks make before-and-after comparisons after a fix easier to read.
• Competitor comparison: the competitor named on each gap row is the input to the Gap factor.
• Citation source detail: knowing which page each citation comes from drives the Source factor.

For a view of how Noma compares with the other tools, see Noma vs AthenaHQ, Noma vs Profound and Noma vs Scrunch. Profound's own comparison page, for example, describes its AI Marketer turning findings into ranked projects with evidence attached (Profound, 2026), which shows that ranked fixes are becoming a standard expectation across the category.

Common mistakes when prioritizing AI visibility fixes

• Ranking by traffic alone. As the worked example shows, a busy page with a small shortfall can outrank a lower-traffic page that costs you the answer.
• Skipping access checks. Rewriting a page that crawlers cannot reach changes nothing.
• Treating one run as truth. With cited sources shifting between modes, a single answer is weak evidence.
• Mixing fix types. Technical, structural and new-content fixes have different effort and different review needs, as Scrunch's three buckets show. Score them separately.
• Fixing without a control. If you do not hold the prompt set constant, you cannot say a fix worked.
• Ignoring third-party sources. When engines cite other sites, the fix sits off your domain. See why a competitor is cited and you are not.

How do you tell whether a fix worked?

Compare the same prompts before and after, on the same engines, over a long enough window to smooth run-to-run variation. Track three things: whether you are named, whether the cited page is the one you changed, and whether the competitor you targeted still holds the slot.

Also connect visibility to visits. AI referrals show up in analytics under different referrers, which our guide on measuring AI referral traffic explains. For a monthly view, use the AI search metrics monthly report structure, and check your starting position with the free GEO readiness scorecard.

Should you hire help to prioritize and ship the fixes?

If your team lacks time to run the checklist, a managed service can carry it. Nurdd's team runs answer engine optimization and generative engine optimization services, measured in Noma. For comparing providers first, read best AEO and GEO agencies in 2026 and agency versus in-house tooling.

For the wider category, best AI search visibility tools in 2026 and what generative engine optimization is give the context behind the method here.

How we researched this

We reviewed Scrunch's public product pages, launch blog, help center and pricing page, plus Profound's public comparison page, Google's published guidance and Semrush research pages. All pages were checked on the publication date of this post. We did not test Scrunch hands-on, and we have no access to its model, so everything stated about how Scrunch ranks fixes comes from what Scrunch publishes.

The sample figures in the worked example are the illustrative numbers shown on Scrunch's Site Optimization page. The scoring framework and the multiplication are our own and are not Scrunch's method. Rival plans and prices can change, so confirm them on the vendors' pages.

Frequently asked questions

How does Scrunch decide which website fixes matter most?

Scrunch says Site Optimization uses a predictive model that weighs a brand's AI search performance, its competitors' performance and Scrunch's own research. Fixes are ranked by impact and mapped to URLs. Scrunch has not published the model's weights, so the exact formula is not public.

How do I prioritize fixes for AI search visibility?

Check crawler access first, then list your tracked prompts, record the competitor named and the source cited on each gap, and score each fix on gap, source, page importance and effort. Ship the top few, then re-test the same prompts. The checklist above sets out each step.

What are the three types of fixes Scrunch describes?

Scrunch groups fixes into technical normalization, content restructuring and content enrichment. Technical changes are auto-approved, restructuring can be approved or auto-approved, and enrichment uses an approve-and-deploy workflow, according to its launch post.

Do I need a tool to rank fixes, or can I do it by hand?

You can do it by hand for a small prompt set using a spreadsheet and the four-factor score. A tool helps once you track many prompts across many engines, because collecting gaps and cited sources manually does not scale. Noma by Nurdd tracks 13 engines and compares competitors for you.

Should crawl and access issues come before content fixes?

Yes. Scrunch's own crawlability FAQ says to start with access controls and then fix the pages that matter most. A page AI crawlers cannot retrieve cannot be cited, so content work on it has no effect until access is repaired.

Does Noma have a free trial?

Noma offers a 3-day free trial. A card is taken at signup and billing starts on day 3. If you cancel before then, you are not charged.

How long before a fix shows up in AI answers?

It varies by engine and by how often the engine re-reads your page, and no source we found gives a reliable timeline. Re-test the same prompts at two and four weeks and compare, rather than relying on a promised date.

Next step: see your own ranked gaps

A ranked list is only useful if it is built from your prompts, your competitors and your cited sources. Book a demo at nurdd.club/demo and we will walk through your brand in Noma: where the 13 engines name you, where they name a competitor instead, and which page each citation comes from. You can also look at Noma directly, or read how Nurdd works.