How accurate is Scrunch data? Collection, refresh and history
A source-checked guide to how Scrunch collects AI answers, detects brand mentions, refreshes prompts, re-evaluates matching rules, and limits dashboard and API history.
Quick answer
How accurate and current is Scrunch data for an AI visibility decision?
Scrunch collects actual platform responses through browser automation and official APIs, then detects explicit brand mentions with pattern matching. New prompts run daily for 14 days before a default 72-hour cadence, with manual refresh available. Matching controls can re-evaluate history. The method is unusually specific, but Scrunch does not publish an independent precision or recall benchmark, so buyers should validate aliases, exclusions, citations, and repeatability on their own prompt set.
Key facts and evidence
- Collection
- Platform-specific browser automation and official APIsEvidence: How Scrunch collects data from AI platforms
- Default refresh
- Daily for 14 days, then every 72 hours, with on-demand refreshEvidence: How Scrunch collects data from AI platforms
- Mention rule
- Explicit matched text; a citation alone does not count as a mentionEvidence: How Scrunch tracks brand presence across LLMs
- Dashboard history
- Up to 12 months or environment creation, whichever is shorterEvidence: Does Scrunch show change over time, Scrunch API products and operation
- API history
- Latest 90 days for Query and Responses dataEvidence: Does Scrunch show change over time, Scrunch API products and operation
How to interpret Scrunch data
| Decision point | What Scrunch documents | What remains uncertain | Buyer check |
|---|---|---|---|
| Answer capture | Actual responses collected through browser automation or official APIs | The public page does not map each surface to one collection method | Ask which method applies to every required platform and geographyEvidence: How Scrunch collects data from AI platforms, How Scrunch tracks brand presence across LLMs, Scrunch Configuration API overviewNot publicly verified. A private audit or customer-specific validation may exist. Buyers should request evidence for their own brand aliases, markets, and required platforms. |
| Brand mentions | Pattern matching against response text, separate from citation detection | No public precision or recall score for ambiguous brand names | Test aliases, case rules, exclusions, and names shared with common wordsEvidence: How Scrunch tracks brand presence across LLMs, Scrunch Configuration API overview, How Scrunch collects data from AI platformsNot publicly verified. A private audit or customer-specific validation may exist. Buyers should request evidence for their own brand aliases, markets, and required platforms. |
| Freshness | Daily for the first 14 days, then a 72-hour default with manual refresh | A manual refresh is not the same as a contracted collection service level | Set the required decision cadence before comparing trend changesEvidence: How Scrunch collects data from AI platforms |
| Historical changes | Matching configuration changes can trigger historical re-evaluation | Reclassified history may not be directly comparable with an older export | Record configuration changes alongside reporting periodsEvidence: Scrunch Configuration API overview |
| Retention | Dashboard history is bounded at 12 months and API history at 90 days | Public material does not promise pre-environment response backfill | Export response-level data if the buying case needs longer analysisEvidence: Does Scrunch show change over time, Scrunch API products and operation |
Checked 25 August 2026. Collection accuracy and visibility represent measured platform responses, not every answer every user could receive.
How does Scrunch collect answers, and does it use the same method everywhere?
Scrunch says it captures actual responses using a platform-specific mix of browser automation and official APIs. That is more concrete than a generic claim that data comes from leading models.
The current public method page does not assign a collection method to each platform. A buyer should request that mapping for required engines, logged-in or public surfaces, locations, and any plan-specific differences.
How does Scrunch decide whether an answer mentions my brand?
Scrunch matches explicit brand text with regular expressions and patterns. It treats citations as a separate signal, so a linked brand domain without the brand name in the answer does not count as a mention.
Case sensitivity, excluded names, aliases, and domain classification give teams ways to reduce false matches. Configuration changes can re-evaluate historical records, so analysts should note when definitions change.
Is Scrunch daily data, or does it refresh less often?
Both descriptions can be true at different points in a prompt's life. Scrunch documents daily collection for the first 14 days, followed by a default refresh every 72 hours. A user can also request a refresh.
For launches, incidents, or short tests, confirm whether on-demand runs are included and how quickly they complete. For steady reporting, compare like-for-like windows rather than treating an individual response as a stable market fact.
How much Scrunch history is available in the dashboard and API?
The dashboard can show up to 12 months or the age of the environment, whichever is shorter. Scrunch's Query and Responses APIs expose the latest 90 days.
That difference matters for warehouse design and vendor migration. A team that needs response-level analysis beyond 90 days should schedule its own extraction instead of assuming the API will return the full dashboard range later.
Evidence and method
Scrunch separates mentions from citations
Its official accuracy page gives a concrete counting rule, which lets buyers reproduce a basic mention check instead of relying on an unexplained visibility label.
Evidence: How Scrunch tracks brand presence across LLMsMatching controls can correct ambiguous names
Case rules, exclusions, aliases, and historical re-evaluation provide practical controls for brands whose names overlap with people, places, or common words.
Evidence: Scrunch Configuration API overviewFreshness and retention have explicit bounds
Scrunch publishes its default prompt cadence and separate dashboard and API history windows, so teams can plan tests and exports around known limits.
Evidence: How Scrunch collects data from AI platforms, Does Scrunch show change over time, Scrunch API products and operationIndependent accuracy validation remains open
The public method is detailed, but no published precision, recall, or cross-platform benchmark establishes how often Scrunch classifications are right for a buyer's own dataset.
Evidence: How Scrunch collects data from AI platforms, How Scrunch tracks brand presence across LLMs, Scrunch Configuration API overviewNot publicly verified. A private audit or customer-specific validation may exist. Buyers should request evidence for their own brand aliases, markets, and required platforms.How we checked this page
We checked Scrunch's current collection, accuracy, configuration, history, and API documentation, separated vendor-described mechanics from independently verified performance, and compared Trakkr from current Trakkr sources.
- 1. Mapped each collection, matching, refresh, and retention statement to the specific official page that defines it.
- 2. Kept citations, mentions, configuration re-evaluation, dashboard history, and API history separate because they answer different accuracy questions.
- 3. Marked the absence of a public validation benchmark as uncertainty rather than evidence that the method fails.
- Limitation: We did not access a Scrunch account, raw response set, private customer contract, or independent classification audit.
- Limitation: AI outputs vary by time, location, account state, and model behavior, so no monitoring sample represents every possible user answer.
When is Scrunch or Trakkr the stronger data fit?
Choose Scrunch when your team values explicit matching controls and wants to re-evaluate history after changing aliases, exclusions, or case rules. Its public method gives analysts more levers for difficult brand names.
Choose Trakkr when a simple, consistent daily cadence across eight named models matters more. Trakkr says every active prompt runs on all eight each day, while Scrunch moves prompts to a default 72-hour cadence after 14 days.
No. Scrunch says the brand name must appear explicitly in the response for a mention to count; a citation to the brand's domain is tracked separately.
Scrunch documents case-sensitive matching, excluded names, aliases, and domain rules. Changes to these controls can trigger historical re-evaluation of affected data.
Not by default. Scrunch says new prompts run daily for 14 days, then move to a 72-hour default cadence, with on-demand refresh available.
We found no current public precision, recall, or cross-platform accuracy benchmark. The collection and matching controls are documented, but buyers still need a test using their own brands and prompts.
Sources and related reading
See how AI talks about your brand
Enter your domain to get a free AI visibility report in under 60 seconds.