"Honesty over hype" is one of the values on our About page. It reads simply on a slide. In practice, it's the value we expect to cost us the most, and it's worth explaining why.
AI marketing has a structural problem. Most buyers can't verify what a vendor claims until after they've bought, so the most confident claim tends to win, whether or not it's true. "Replaces your entire front desk." "100% accuracy, guaranteed." "Understands your customers like a human would." Those lines out-convert careful ones almost by construction.
We don't write lines like those. We like marketing fine; we just think the bill for hype always comes due, usually a few weeks after someone signs up, when the product meets the promise.
If you're buying AI software rather than selling it, the four checks near the end are the part to keep, and the rest of the essay explains why they're necessary.
Why AI marketing rewards confidence over accuracy
Verification happens after the purchase
You can't test-drive "understands your customers." You can't kick the tyres on "99% accurate." In most AI software, the only real test is putting it to work on your own business, which happens after the decision.
So in the moment of choosing, the claim that sounds most confident tends to win the click. Individual buyers aren't being careless; the pattern comes from a category where proof arrives late.
Puffery versus checkable claims
There's a line between vague enthusiasm and a specific, testable promise. Advertising standards bodies such as India's Advertising Standards Council exist to scrutinise claims once they cross into checkable territory: a guarantee, a percentage, a head-to-head comparison.
US regulators pay attention too. The FTC's artificial intelligence page lists enforcement actions over unsubstantiated AI claims, including a company ordered to stop overstating how accurate its AI detection tool was. Most AI copy is written to stay on the safe, vague side of that line, which makes it hard to challenge and easy to keep writing.
What honest AI marketing looks like
Claims we don't make
We don't say Aiva Chat handles every call. We don't promise 100% accuracy. We don't say it replaces your team. We don't publish resolution percentages, customer counts, or conversation volumes, because we're an early-stage company and any such number would either be invented or borrowed from somewhere it doesn't belong.
What we say instead
We say Aiva Chat answers questions from your own information, books appointments, qualifies leads, and hands off the calls it shouldn't try to handle. We say it needs a few weeks of watching and adjusting before it fits your business well. We say every business sees its own numbers, including how many calls went to a person.
Each of those is a weaker sentence than the hype version, in the narrow sense of being less exciting to read. We think they're stronger in the sense that matters: they're still true three months later.
Hype vs. honest claims, side by side
| Hype version | Honest version | Why the honest one holds up |
|---|---|---|
| "Replaces your front desk" | "Handles the repetitive calls and hands off the rest" | Matches what owners see in week three |
| "100% accurate" | "Resolves what it can and shows you what it handed off" | The handoffs are visible, not hidden |
| "Works out of the box" | "Plan on a few weeks of reviewing calls and adjusting" | Sets up the work that makes it good |
| "Understands customers like a human" | "Follows natural conversation and knows when to get a person" | Testable on a single call |
| "Thousands of happy businesses" | No customer claims at all | We're early-stage and say so |
| "Never misses a call" | "Answers every call it receives, within your plan's limits" | Concurrent-call limits are real |
To be fair to the hype column, it isn't always written in bad faith. Sometimes it's aspiration, sometimes it's a best-case demo stretched into a headline. The problem is that buyers can't tell which, and they find out on their own time and money.
The honest column has its own weakness: it's harder to sell in a quick comparison. A buyer scanning two sheets side by side will often see our claims as smaller, and some will choose the bigger-sounding one. That's the cost we're choosing to accept.
The case for rounding up anyway
It's worth steelmanning the other side, because the argument for a little exaggeration isn't purely cynical.
Here's the theory: everyone knows vendors round up, so buyers apply a mental discount to every claim. "100% accurate" gets silently downgraded to something like 85%. A vendor who plainly states a lower, real number just looks worse after everyone's discount is applied. Under that theory, an honest number simply reads as a weaker product.
We think that theory used to be closer to true than it is now. A buyer burned once by a "100% accurate" claim doesn't discount the second time; they stop trusting the number at all and look for a vendor willing to say something checkable. We're betting on that shift. It's a bet, not a proven result.
What honesty costs us
The comparison-sheet problem
Picture it: a buyer lines up a competitor's sheet next to ours and asks why our claims look smaller. The honest answer, that big numbers are often aspirational and the number that counts is the one your own account produces, doesn't always land in a five-minute comparison. Sometimes the bigger sheet wins.
Why we accept the loss
That lost sale is often a preview of a worse conversation we'd have had three months later, when the product didn't match what the buyer expected. It just happens earlier, on someone else's desk.
Hype is a loan against future trust. It always comes due, usually in week three, when the product meets the promise.
Making AI claims checkable instead of bigger
If we're not going to out-promise competitors, the honest alternative is to make our claims checkable.
Explaining what our numbers measure
We publish what a resolution-rate number actually measures, including how a single average can hide a badly performing category underneath it. A number you can interrogate is worth more than a headline you have to trust.
Letting owners hear their own calls
Every call is saved with its recording, transcript, and summary. Owners can listen to what their assistant actually said instead of relying on a dashboard summary. That's slower than a confident claim, and it's the thing that holds up when a sceptical buyer checks.
Showing the handoffs
The dashboard shows calls resolved without a person, which means it also shows the ones that weren't. We'd rather owners see that split clearly than discover it by accident.
Why a bootstrapped company can afford this
Aiva Chat is bootstrapped. We've never raised outside capital, so no investor is judging us on a quarter-over-quarter curve. That isn't a moral advantage. Plenty of funded companies are honest, and plenty of bootstrapped ones hype shamelessly.
But it does remove one specific pressure: the temptation to inflate a claim to hit a growth number on someone else's slide. The growth number that matters to us is whether people keep using the product, and that doesn't respond well to promises that don't hold up.
It also shapes small decisions. We describe setup as six guided steps rather than quoting a number of minutes, because how long it takes depends on how much you want to teach the assistant. We say every feature is on every plan, because that's true, rather than inventing tiers to make the top one look generous. None of these choices is dramatic on its own. Together they add up to marketing a buyer can check line by line.
How to spot AI hype when you're buying
If you're evaluating AI software for your business, a few checks cut through most of it.
Step 1: Ask what each number measures
When a vendor quotes a percentage, ask what it counts, over what period, and whose data it comes from. If the answer is vague, treat the number as marketing.
Step 2: Insist on a live call
Recorded demos are edited. Call the product yourself, interrupt it, ask two things at once, and ask something it shouldn't know.
Step 3: Ask what happens when it fails
Every AI system fails sometimes. A good vendor will describe exactly how theirs hands off. If they claim it never needs to, be wary.
Step 4: Trial it on your own calls
Your callers, your questions, your accents. A free trial on real traffic is the only test that matters. Watch the calls it hands off as closely as the ones it resolves, because that's where you learn whether the vendor's description was honest.
Common mistakes in AI marketing
Quoting other people's results as yours
A resolution rate from one business tells you little about another with different callers and information. Presenting it as a product property is misleading even when the number is real.
Treating a demo as a guarantee
A perfect scripted demo shows the best case. Marketing that implies every call will go that way sets buyers up to feel misled.
Inventing social proof
Testimonials that aren't real, or "trusted by thousands" from a company with a handful of users, are the fastest way to lose trust once someone looks closely.
Hiding the human handoff
Vendors sometimes treat handoffs as a weakness to hide, when a clean handoff is one of the most useful things the product does. Pretending the AI never needs a person makes the first transfer feel like a failure.
Principles we hold ourselves to
- No claim we can't defend if a customer asks for proof.
- No borrowed numbers. Your dashboard, your data.
- No testimonials or customer counts until they're real and verifiable.
- Say what it doesn't do, such as no payments taken inside a text conversation and no deposits taken during a phone call.
- Show the handoffs, not just the resolutions.
- Describe setup honestly: six guided steps, then weeks of tuning.
- Avoid empty labels, which is also why we don't say "AI-powered".
Conclusion
Hype is a rational short-term strategy in AI marketing, because buyers can't check claims until after they've paid. We've decided against it, knowing it will sometimes cost us a comparison against a bigger-sounding sheet. Many small business owners already fear being oversold on AI, and we'd rather earn their trust slowly than borrow it.
The caveat is that this is a bet. Our guess is that hype's shelf life is shortening as more buyers get burned, but we don't know how fast, and we may lose deals to it in the meantime.
For a buyer, the practical takeaway is simpler than the argument: treat every number you can't check on your own calls as marketing, ours included.
Ready to check our claims yourself?
Hold us to the checks in this essay. Set Aiva Chat up on your own business during the 14-day trial (no card needed), then read the transcripts of the calls it handed off as closely as the ones it resolved.
If it holds up, plans are $49, $99, or $199 a month with every feature on each, and your OpenAI and Twilio accounts carry the usage. Read more about who we are and how we work, or call +1 (970) 972-0974 and try to catch it overpromising.
Start your free trial, or book a demo and bring the vendor claim you trust least.
Posts from the people building Aiva Chat in Rajkot — how the product works, what we have learned building it, and guides for getting the most out of an AI receptionist.