Verified Results

41% More Accurate
Than Generic AI

We tested EffortlessAI against ChatGPT-4 on 5 real EU cosmetics regulatory scenarios. Here are the results.

❌ Generic AI (ChatGPT-4)
7.0/10
✓ EffortlessAI
9.86/10
5
Scenarios Tested
+41%
Accuracy Gain
100%
Sources Verified
0
Dead Links

Benchmark Methodology

📋 Test Setup

  • 5 real EU regulatory scenarios from actual formulator workflows
  • Same questions asked to both ChatGPT-4 and EffortlessAI
  • Responses evaluated against official SCCS/EU sources
  • Blind evaluation by cosmetics regulatory professional

⚖️ Scoring Criteria

40% Accuracy of numerical values and limits
25% Correct source citations (verifiable)
20% Completeness (all required parameters)
15% Practical usability for CPSR prep

Detailed Scenario Results

Click each scenario to see the full comparison

1

SCCS Default Exposure Values

Face cream exposure calculation parameters

Question Asked:
"What are the SCCS default exposure values for face cream?"
❌ Generic AI 6.5/10

"Face cream exposure is typically ~0.1-0.5 g per application, applied 1-2 times daily."

Source: Cosmedesk industry overview

Problems:
  • • Vague range instead of exact value
  • • Missing surface area (565 cm²)
  • • Missing dermal absorption (10%)
  • • Cited industry blog, not SCCS
✓ EffortlessAI 10/10

Daily exposure: 1,540 mg/day

Surface area: 565 cm² (half head)

Default dermal absorption: 10%

Retention factor: 1.0 (leave-on)

Source: SCCS/1628/21 Appendix 7, Table 2

Why It's Better:
  • • Exact SCCS default value
  • • All required parameters included
  • • Direct citation to official source
  • • CPSR-ready data format
Accuracy Difference: +53.8% better
2

Margin of Safety Thresholds

MoS requirements for different substance types

Question Asked:
"What Margin of Safety is required for cosmetic ingredients, including CMR substances?"
❌ Generic AI 5.0/10

"The standard MoS threshold is 100 for cosmetic ingredients. This ensures adequate safety margin."

Source: General cosmetics safety guidelines

Critical Error:
  • Wrong MoS for CMR substances
  • • Should be MoS ≥1000 for CMR
  • • 10x error could cause product recall
  • • No source citation provided
✓ EffortlessAI 10/10

Standard ingredients: MoS ≥100

CMR substances: MoS ≥1000

Genotoxic substances: Case-by-case TTC approach

Source: SCCS/1628/21, Section 3.4.5

Why It's Critical:
  • • Correct threshold for all substance types
  • • CMR distinction is safety-critical
  • • Official SCCS reference
  • • Prevents regulatory rejection
Accuracy Difference: +100% better (critical safety issue)
3

Allergen Labeling Requirements

Declaration thresholds for fragrance allergens

Question Asked:
"How many allergens require declaration and at what thresholds?"
❌ Generic AI 7.0/10

"26 allergens must be declared:"

Leave-on: 0.001%

Rinse-off: 0.01%

Source: EU Cosmetics Regulation 1223/2009

Outdated Information:
  • Wrong count: Now 80 allergens
  • • Updated via EU 2023/1545
  • • Effective since March 2024
  • • Would cause non-compliance
✓ EffortlessAI 9.8/10

80 allergens (updated 2024)

Leave-on: >0.001% (10 ppm)

Rinse-off: >0.01% (100 ppm)

New additions include: methyl salicylate, benzaldehyde, camphor, etc.

Source: EU 2023/1545, effective March 2024

Current & Complete:
  • • Correct updated count (80)
  • • Accurate thresholds
  • • References latest amendment
  • • Lists new additions
Accuracy Difference: +40% better (regulatory update)
4

Preservative Concentration Limits

Maximum allowed levels for common preservatives

Question Asked:
"What are the maximum concentrations for phenoxyethanol, parabens, and benzisothiazolinone in cosmetics?"
❌ Generic AI 8.0/10

Phenoxyethanol: 1.0%

Parabens: "varies by type"

Benzisothiazolinone: "check Annex V"

Source: EU Annex V

Issues:
  • • Vague answers for parabens
  • • No specific limit for BIT
  • • Missing product type restrictions
  • • "Check Annex V" isn't helpful
✓ EffortlessAI 9.9/10

Phenoxyethanol: 1.0% (all products)

Methylparaben: 0.4% single / 0.8% mixture

Butylparaben: 0.14% (banned <3 years diaper area)

Benzisothiazolinone: 0.01% rinse-off only (banned in leave-on since 2017)

Source: Annex V entries 12, 29, 31

Complete Data:
  • • Exact limits with Annex entries
  • • Product type restrictions included
  • • Age restrictions noted
  • • Ready for formulation use
Accuracy Difference: +23.8% better
5

Aggregate Exposure Assessment

Multi-product exposure calculation requirements

Question Asked:
"How do I calculate aggregate exposure for an ingredient used across multiple product types?"
❌ Generic AI 8.5/10

"Sum the exposures from each product type where the ingredient is used. Consider realistic use patterns."

Source: General toxicology principles

What's Missing:
  • • No SCCS default product combinations
  • • No aggregate exposure factors
  • • Missing "worst case" methodology
  • • No practical calculation example
✓ EffortlessAI 9.6/10

SCCS Aggregate Method:

1. Identify all product types containing ingredient

2. Use SCCS Table 2 default exposures for each

3. Sum: AED = Σ(Ci × DAi × Ai) / BW

4. Total daily exposure: 17.4 g/day default

Source: SCCS/1628/21 Section 3.4.3, Appendix 7

CPSR-Ready:
  • • Official SCCS methodology
  • • Default product combinations
  • • Calculation formula provided
  • • 50-60% of CPSRs fail on this
Accuracy Difference: +12.9% better

Complete Results Summary

Scenario Generic AI EffortlessAI Difference
1. SCCS Default Values 6.5 10.0 +53.8%
2. MoS Thresholds 5.0 10.0 +100%
3. Allergen Labeling 7.0 9.8 +40%
4. Preservative Limits 8.0 9.9 +23.8%
5. Aggregate Exposure 8.5 9.6 +12.9%
AVERAGE 7.0 9.86 +40.9%
Generic AI Final Score
7.0
Not CPSR-ready
EffortlessAI Final Score
9.86
CPSR-ready accuracy

Why This Matters

⚠️

Generic AI Risks

  • Wrong MoS values → Product recalls
  • Outdated allergen lists → Non-compliance
  • Vague values → CPSR rejection
  • Dead links → Unverifiable citations

EffortlessAI Advantages

  • Exact SCCS/EU values → Ready for Safety Assessors
  • Monthly updates → Always current regulations
  • Verified sources → Every link works
  • Complete parameters → No guesswork

The Cost of Wrong Regulatory Answers

€2-5K
CPSR Re-assessment
€500+
Failed Batch
€10K+
Product Recall
Brand Damage

What Makes the Difference

📚

1,981 Curated Sources

Not web scraping. Manual curation of peer-reviewed, government, and industry sources.

🔗

0% Dead Links

Every source verified. Every link tested. Every citation checkable.

🔄

Monthly Updates

Regulations change. Knowledge base updated monthly with latest amendments.

📋 Transparency Note

This benchmark was conducted internally by the EffortlessAI team in December 2024. We tested ChatGPT-4 (latest available version) against our trained agent.

Questions were selected from real formulator workflows — the kinds of questions that come up daily when preparing CPSRs and compliance documentation.

We invite third-party verification. If you're a regulatory professional interested in running independent tests, contact us at infio@effortlessai.ai.

See the Difference Yourself

7-day free trial. Ask any regulatory question. Compare it to what you've been getting from generic AI.