Front Desk · 2026-08-15

What You Can and Cannot Claim About an AI Receptionist in the UK

Why this matters now

Any business marketing AI software in the UK — and any small business repeating a vendor's performance claims on its own site — falls under the rules enforced by the Advertising Standards Authority (ASA) and the Committee of Advertising Practice (CAP). The codes apply equally to a SaaS company's landing page and to a salon owner's Instagram ad that quotes their chatbot vendor's statistics.

The consequences of getting it wrong are real: formal complaints, mandatory ad bans, and the reputational damage of a published ASA ruling against your business. This is a practical guide to what the rules require, with examples drawn from the speed-to-lead evidence base that underpins most AI receptionist marketing.

The three risk categories

Absolute claims

Statements like "never miss a lead" or "guaranteed 100% conversion" are heavily scrutinised under CAP codes. The ASA operates on the principle that absolute claims must be supported by robust, real-world data. No AI system can guarantee absolute performance under all conditions — technical failures, network outages, and API limitations make "never" and "guaranteed" indefensible without extraordinary evidence.

The ASA has banned ads for making performance claims that could not withstand testing (a recent ruling targeted portable cooling devices claiming to cool a room in 90 seconds without scientific evidence). The same logic applies to software: if you cannot prove it with production logs covering edge cases, do not claim it.

Comparative claims

Claiming your software outperforms competitors — "faster than a human receptionist" or "the best WhatsApp automation on the market" — requires verifiable comparative evidence using objective metrics. You must keep the testing data accessible for public or regulatory inspection. Vague comparisons ("better than the alternatives") fail this test. Specific, evidenced ones ("our median automated response time is 6 seconds, compared to the industry average of 12 hours 10 minutes for manual email replies reported by Ringly.io and thisandthat.chat") are defensible, provided the comparison is like-for-like and the data is retained.

AI-washing

Claiming AI-driven capabilities that do not map to demonstrable technical functions risks an "AI-washing" complaint. If your marketing says the system "understands context and nuance," you need to be able to show what that means in practice — not just that it uses a language model.

What is safe to say

The evidence report used across this blog series grades every statistic by provenance. Here is how that grading translates into compliant copy:

Fully substantiated (safe): - "Answer incoming WhatsApp enquiries 24/7 without hiring extra front-desk staff." This is a factual description of what automated API webhooks do. - "Industry data shows that responding under five minutes can convert up to 32% of enquiries, compared to just 12% when delayed past 24 hours." This directly cites the Optifai benchmark covering Q2 2025 to Q1 2026 across 939 B2B SaaS companies — grade (b), large-sample vendor benchmark. - "Keep a warm, responsive presence during evening class hours and weekends when the front desk is busy." This aligns with documented after-hours enquiry arrival patterns (Dance Business Weekly, 2025; Blazeo, 2026).

Requires local data to be safe: - "Respond to parents in seconds, 24 hours a day." To use "in seconds," you need API timestamp logs showing the median automated response is delivered in under 60 seconds. If the median is 6 seconds, the claim is defensible. If you have not measured it, do not say it.

Unsafe (subject to ASA action): - "Our AI ensures your business will never miss a lead again." - "Guarantees a 300% increase in student enrolments." - "The fastest customer response system in the UK."

Citing statistics responsibly

The speed-to-lead literature is saturated with recycled numbers that circulate without dates or methodological context. Two rules keep your marketing honest:

Carry the vintage. Where a statistic has a data year, state it. The "100x contact rate" and "21x qualification multiplier" come from a 2007 study by Dr James Oldroyd at MIT, with retrospectives in 2011 and 2012. Quoting them as current fact, without dates, misleads the reader. Saying "a 2007 study of over 15,000 B2B leads found a 100x difference in contact rates for five-minute responses — and the underlying pattern was confirmed by the Optifai benchmark covering Q2 2025 to Q1 2026 across 939 B2B SaaS companies" is honest and still compelling.

Name the source class. A peer-reviewed academic study (grade a) carries different weight from a vendor-sponsored benchmark (grade b), which in turn is different from an unverifiable marketing claim (grade c). The 78% "first responder wins" figure, for example, originates from a self-reported consumer survey by the vendor Lead Connect (circa 2018–2020) with no academic replication. You can reference it as a consumer perception data point. You cannot present it as proven fact.

The substantiation checklist for any landing page:

1. Never cite a metric without naming the underlying research. "Studies show" without attribution is exactly what regulators look for.
2. If claiming "responds in seconds," compile and retain a continuous log of inbound webhook triggers to outbound API replies.
3. Add a clear disclaimer on pricing or performance pages: "Conversion projections are based on standard industry benchmarks and are illustrative of potential operational improvements; individual business results will vary."

Frequently Asked Questions

Do ASA rules apply to a small business quoting its vendor's claims?

Yes. If you publish a performance claim on your website, social media, or advertising — even if you are repeating what your software provider told you — you are the advertiser and you are responsible for substantiation under CAP codes.

Can I say my AI receptionist is better than hiring a human?

Only if the comparison is specific, measurable, and backed by data you can produce on request. "Replies faster than a part-time receptionist" is a comparative claim that needs evidence. "Handles WhatsApp enquiries around the clock without additional staff costs" is a factual description of the operational model and is safe.

What happens if the ASA upholds a complaint against my ad?

An upheld complaint can result in a published ASA ruling, an ad ban, and reputational damage.