01 Independent chatbot audits
Mystery shopping for chatbots, with real evidence behind every finding.
DECOY reveals where the experience fails and why.
No SDK. No backend access. We test the live customer experience.
How a DECOY finding reads: what the customer saw, what it costs, what to change.
DECOY · ISSUE RECORDED
HIGH · SYSTEMIC · REVENUE
The customer disclosed team size, timing, and purchase intent. The chatbot never offered a demo, captured their details, or handed them to sales.
A qualified lead can leave without entering the sales pipeline.
DECOY · ISSUE RECORDED
CRITICAL · CONVERSION · REVENUE
The customer asked a simple product question, was gated behind an email request, and asked again after the bot admitted the gate was unnecessary. They named a competitor and ended the conversation.
This is not a hypothetical lost sale. The customer said, in the transcript, exactly where it went.
DECOY · ISSUE RECORDED
CRITICAL · TRUST · ACCURACY
The customer pointed directly at where the information was published. The chatbot repeated the same false claim rather than checking or correcting itself.
This wasn’t a guess. It was a published fact, denied twice, on the record.
What you receive
Aggregate audit report
Journey reports and verbatim transcripts
Prioritised findings and recommended actions
Customer Experience Floor: your worst visit, next to your average
02 The blind spot
Your analytics show that the chatbot responded - not whether it understood, answered correctly, converted the lead, or knew when to escalate.
DECOY reveals the business consequences your dashboard cannot see.
Illustrative comparison: not client performance data
03 Business consequence
Three executive lenses. One customer experience.
R / 01
Does your chatbot recognise commercially valuable visitors and move them toward a purchase, booking, demo, or qualified conversation?
A / 02
Are the answers correct, consistent with published information, and free from invented policies, prices, or commitments?
T / 03
Does your chatbot handle objections, uncertainty, context, and human escalation without frustrating or misleading the customer? And does it tell customers they are talking to an AI, as the EU AI Act has required since August 2026?
04 Journey performance
As customers move closer to a purchase or resolution, their intent changes. The audit follows that evolution across four end-to-end stages.
01
Can an undecided visitor understand the options?
02
Can the chatbot make and explain a useful recommendation?
03
Can it recognise purchase intent and progress the opportunity?
04
Can it resolve the issue or escalate cleanly to a human?
Illustrative report format. Scores do not represent a named client.
05 Evidence standard
DECOY’s core analytical value is the transcript-to-finding specimen: a structured path from observation and evidence to consequence, action, and verification.
DECOY · ISSUE RECORDED
Thank you, This enterprise plan looks right for our team. Can you walk me through onboarding and next steps?
Our Enterprise plan is very popular. You can learn more in the Help Centre. Is there anything else I can help with?
HIGH · SYSTEMIC · REVENUE
The customer selected a tier and asked for next steps. The chatbot returned a passive link and attempted to close the conversation.
06 Signature metric
Aggregate performance can look healthy while high-value customers experience critical failures. DECOY captures both.
Illustrative DECOY report treatment
07 Has your chatbot gone rogue?
DECOY cross-checks chatbot statements against published client information and separates what is verified from what is contradicted or cannot be substantiated.
€89 / month
€59 / month
/pricing
CONTRADICTED
48 hours
48 hours
/terms §4.2
VERIFIED
99.9%
No public SLA
UNVERIFIABLE
Illustrative examples only. Real findings cite client-specific authoritative sources.
08 Before / after proof
Every case study is anchored in controlled evidence, with a clear distinction between observed results and outcomes not yet validated by an external Re-Audit.
Anonymised case 02, pet services. One customer journey, audited, fixed and re-audited under identical conditions.
BEFORE / AUDIT
AUDITED 18 AUG 2026
INTERVENTION / OPTIMIZE
CHANGE SHIPPED AUG 2026
AFTER / RE-AUDIT
MEASURED 26 AUG 2026
One journey, re-run once: the evidence is the answer the customer got, not the headline number. Lead handling only moved from 78 to 85, because the fix did not target it.
09 Method
DECOY is a specialist service - not software your team has to install, configure, or learn.
Controlled synthetic personas tailored to your products, buyers, and known commercial concerns.
Custom generated personas complete multi-turn journeys in the public customer interface, just like a real visitor.
Responses are scored, reviewed, and checked against authoritative published information.
Leadership sees the verdict & delivery teams get transcript evidence and verification conditions
30 seconds Executive verdict
5 minutes Findings and priorities
30 minutes Evidence and methodology
10 Ways to work with DECOY
Controlled synthetic customers across four end-to-end journey stages, tailored personas, factual verification, an AI disclosure check (EU AI Act), systemic findings, Customer Experience Floor analysis, actionable recommendations, and an executive findings presentation. Run and reported in your chatbot's language.
Findings within 2 to 3 working days of kickoff.
Request a quote ↗A focused engagement that converts Audit findings into specific behavioural changes and practical fixes, then verifies their impact through Re-Audit.
Periodic outside-in testing against an established baseline to detect regressions after pricing, content, model, knowledge-base, or platform changes.
The AI disclosure check covers what customers see: whether the chatbot says it is an AI when a customer arrives, and when asked directly. It is a check, not legal advice.
AUDIT → OPTIMIZE → RE-AUDIT → ASSURANCE
Founder-led specialist
DECOY is led by Henri Blanc-Francard, a creative technologist with more than fifteen years working with luxury, fashion, design and entertainment brands, in a world where every detail of the customer journey, from tone of voice to technical execution, is expected to reflect the brand. Since 2021 that work has moved into conversational AI, he is now a Claude Certified Developer.
CLAUDE CERTIFIED DEVELOPER15+ YEARS’ EXPERIENCEBILINGUAL ENGLISH AND FRENCH
11 Non-invasive preflight
Share the website URL. We’ll manually review whether the live experience is suitable for controlled testing. No messages are sent during this check.
Synthetic customer testing evaluates controlled, multi-turn sample journeys across key funnel stages rather than every possible conversational permutation. It provides an empirical diagnostic baseline without replacing internal telemetry.
DECOY audits only systems the client is authorized to test. Paid work requires verified business ownership. Compatibility checks do not send messages to the chatbot. The audit uses synthetic customers only, so it never touches your customers' personal data.