A score with nothing to compare it to is just a number

Tell a business owner their website scores 71 out of 100 for AI readiness and watch what happens. There’s a pause. Then, almost every time, the same question: “Is that… good?” It’s the right question, and a score on its own can’t answer it. If everyone in your industry sits at 45, you’re the one AI reaches for. If your closest competitor is at 88, you’re the one it quietly skips. Same number, opposite conclusion.

Absolute scores lie by omission

This is the oldest idea in business measurement, and it keeps getting rediscovered: you don’t need a good score, you need a better one than the person you’re competing with. Search was always relative. Nobody ever won by ranking “well” — they won by ranking above someone. AI search is no different, except the stakes are sharper, because an AI answer names two or three businesses rather than listing ten.

Here’s the trap with any single-site audit, ours included. You run a scan, you get a number and a list of things that could be better. Every finding is technically true. But the list has no sense of urgency, because nothing tells you which gaps are the ones actually costing you the answer. So the list gets filed and nothing happens. That isn’t a motivation problem — it’s a context problem.

Run the same scan next to your nearest competitor and the list reorders itself. Findings where you’re already ahead stop being urgent. Findings where they’re comfortably ahead become the whole job. You haven’t gathered a single new fact about your own site. You’ve just made the existing facts legible.

Benchmarking doesn’t tell you what’s wrong. It tells you what matters.

What a competitor gives you that a target can’t

A realistic bar. Best-practice guides describe a perfect site nobody has. Your competitor is a real business in your market with your constraints — a budget, a CMS they can’t easily leave, a website somebody built in a hurry three years ago. If they’ve solved something, it’s solvable. That’s far more useful than an ideal.

A read on your category. If both of you are weak in the same place, that isn’t a personal failure — it’s an open lane. Some of the most valuable findings are the ones where neither of you is doing well, because that’s uncontested ground. You cannot see uncontested ground from inside your own report.

Something that actually gets acted on. Uncomfortable but true: “we could improve our structured data” gets nodded at in a meeting. “They’re ahead of us on three of the five things that decide whether AI recommends us” gets a decision. Competitive information travels through an organisation in a way self-assessment simply doesn’t.

What Head-to-Head does

You give it two website addresses — yours and a competitor’s. That’s the whole input. No card, no signup, no email wall, and you see the complete signal-by-signal result rather than a teaser with the interesting part blurred out.

Both sites are then scanned at the same time, with the same checks, on the same model. That parallel bit isn’t a detail — it’s the entire basis for the comparison being worth anything. Two audits run a week apart, or by two tools with two different opinions of what matters, produce two numbers that cannot honestly be subtracted from each other.

What comes back is an overall score for each side, then a verdict on every individual signal: you lead, they lead, or even. Anything within three points reads as even, because pretending a two-point gap is a victory would be silly. The signals group into five areas: GEO (AI) Readiness (can engines reach, read and understand you), SEO (the classic foundations, which still feed AI answers), Trust & Integrity, Multilingual (whether you exist for customers who don’t ask in English) and Agent-Commerce (whether an AI agent could actually transact with you). Every row shows both scores — not a badge, the actual numbers, so you can see whether you’re behind by two points or thirty.

When a site genuinely can’t be measured — some block automated readers, others render nothing until JavaScript runs — we mark those signals “not comparable” and say why. A comparison tool that always finds you a victory is a compliment, not a measurement.

Two things people ask immediately: no, the other site is never notified, and result pages are never indexed by search engines. Comparisons are private to you.

What it deliberately does not tell you

This distinction is the one most likely to get blurred by everyone selling in this space, so we’ll be precise about it. Head-to-Head measures readiness — how prepared each site is for AI engines: whether they can reach it, read it, understand what the business does, and trust it.

It does not tell you whether ChatGPT, Gemini or Copilot actually name you when a customer asks a real question. That’s a different question with a different method — you have to put those questions to the engines and read what comes back. Readiness is what a site controls on its own pages; being cited is earned over months, out in the wider web. Both matter, they aren’t the same thing, and a tool that quietly implies otherwise is one to be suspicious of.

When you want to know who AI actually recommends

That’s the Full Scan, and it’s the natural next question once a comparison has told you where you stand structurally. It runs real queries across Google AI Overviews, ChatGPT, Copilot and Gemini and reports what each one said: how every engine describes and ranks you, your mention rate per engine with a confidence range, the actual answers word-for-word with the sources each engine cited, and — the part most people open first — the competitors AI recommends instead of you. Not the competitor you typed in. The ones the AI chose on its own, which is frequently not who you expected.

It’s still a diagnosis. A much deeper one, across multiple pages and the sources AI trusts, with every finding prioritised in plain language — but the what, not the how.

When you want the fixes made

The Launch Pack is the only tier that gives you the how. Instead of a list of problems you get the actual artefacts: structured data, llms.txt, crawler-access configuration, answer-shaped content and FAQs, a content piece in a South African language, conversion and credibility material — plus a step-by-step implementation guide written for the platform you’re actually on. Apply it yourself, hand it to your developer, or we implement it. It comes with a dashboard to re-scan and a before-and-after view so you can watch the score move.

If you’ve already paid for a Full Scan on the same site in the last 30 days, we credit it toward the Pack. The Launch Pack includes a Full Scan anyway, and we’re not going to charge you twice for the same scan.

One honest caveat

None of this guarantees you get cited — be wary of anyone who tells you otherwise. AI decides what to cite and no vendor controls that. What you can control is whether you’re eligible and ready: reachable, readable, understandable, trustworthy. Technical readiness moves fast, often within days. Being consistently recommended is a slower, compounding effort measured in months.

But you can’t work on a gap you can’t see, and you can’t tell whether a gap matters without something to measure it against. Start with the free comparison, and pick the competitor you’d least like to lose to.