Most AI tool directories are lists. We're a publication. That means we don't try to be comprehensive — we try to be useful. The question we're trying to answer for every Australian who lands on this site is the one most directories duck: "Should I use this, and what should I pay for it?"
This page explains how we get to that answer. If you've ever wondered why one tool got 22 out of 25 and another got 19, or how we decide which tools to cover at all, this is the page that shows the working.
What we cover, and what we don't
We review AI tools that are available to Australians, in active development, and useful to the audience we write for: small business owners, professionals, founders, and everyday Australians who want to use AI without becoming a developer.
That means we deliberately don't cover:
- Tools in closed beta with no public access
- Pure developer infrastructure (model APIs, vector databases, fine-tuning frameworks) — there are better publications for that
- Tools that aren't available in Australia, unless there's a strong reason to flag the absence
- Anything we can't trial or verify against solid primary sources — if we can't stand behind the facts, we don't publish
We pick tools based on what we'd want our friends, clients, and family in Australian small business to know about — not what's trending on US tech Twitter.
The five-criteria scoring framework
Every full review on the directory is scored on five criteria, each out of five points, for a maximum of twenty-five. The criteria are deliberately weighted equally. We're not trying to hide editorial judgement behind a complicated formula — we're trying to make sure every review covers the same ground.
1. Value for Money (out of 5)
Is the price fair for what you actually get? Would an Australian user feel ripped off or pleasantly surprised?
A tool gets a five if the free tier alone is genuinely useful, or if paid plans are competitively priced in AUD. It gets a one if it's overpriced, locks core features behind high-tier plans, or sells in USD without flagging GST.
2. Ease of Use (out of 5)
How quickly can a non-technical Australian get started and see results?
A five means useful within minutes, no training required. A one means confusing, buggy, or requires a developer to get value at all. Where we can trial the tool we judge this hands-on — if writing our first prompt takes an hour of docs, that's a real signal. Where we can't, we weigh onboarding from the vendor's own documentation and walkthroughs.
3. Feature Depth (out of 5)
Does it genuinely deliver on its core promise? Does it go beyond the basics?
A five means best-in-class — sets the standard for the category. A one means it doesn't deliver on what it advertises. We're not impressed by feature lists; we're looking at whether the things the tool claims to do actually work, reliably, on real Australian use cases.
4. Australian Fit (out of 5)
How well does this tool work for Australians specifically?
This is where we differ from US-built directories. A five means AUD pricing, local data residency, AU-based support, and outputs that don't get confused by Australian context. A one means the tool isn't available here, raises serious data concerns, or treats us as an afterthought.
If a tool prices in USD only, that's an automatic point off — not because it's a fatal flaw, but because exchange-rate variability and GST add real friction for Australian buyers. If a tool stores data in the US only, we say so. If it doesn't know what a BAS is, we note it.
Added 20 September 2026 — the acquisition path counts too. Since 9 September 2026, Australian app stores must age-check downloads of apps rated R18+, and since 9 March 2026 eSafety's Age-Restricted Material Codes have required an AI chatbot capable of generating sexually explicit, self-harm or high-impact violence material to confirm the user is 18 or older. So we now ask a further Australian Fit question of every mobile-first AI app: what does the store make you prove before you can install it, and who holds that proof? We say in the review what we find. This does not add a sixth criterion. It sharpens the fourth.
Australian Fit leads our verdict. The five criteria are scored equally — but Australian Fit is the reason this publication exists, so it's the criterion we put front and centre. Every verdict explicitly addresses how a tool stacks up for Australians before anything else, and where a tool is strong everywhere but weak on local fit, we say so plainly rather than letting a high overall score paper over it. A globally excellent tool with no AUD pricing, no local data residency, and no AU support is still a tool you should think twice about — and we'll tell you that even when the other four criteria score well.
5. Reliability & Support (out of 5)
Can you depend on it? What happens when something goes wrong?
A five means rock solid uptime, responsive support, active community or documentation. A one means unstable, abandoned, or impossible to get help from. We weight this more heavily than most directories because for any Australian using AI in a business workflow, "the tool was down for six hours and there's no support number" is a real cost.
How the total maps to a star rating
Five criteria × five points = twenty-five points total. We translate that to a star rating because most readers want a quick visual signal:
| Total | Stars | What it means |
|---|---|---|
| 21–25 | ⭐⭐⭐⭐⭐ | Excellent — best-in-class for its category |
| 16–20 | ⭐⭐⭐⭐ | Very good — recommended with minor caveats |
| 11–15 | ⭐⭐⭐ | Good — fit-for-purpose if it matches your specific use case |
| 6–10 | ⭐⭐ | Below average — better options usually exist |
| 1–5 | ⭐ | Poor — avoid |
We don't grade on a curve. A tool that scores 14 out of 25 doesn't suddenly become a four-star pick because everything else in the category scored worse.
The Australian Fit gate
Australian Fit isn't just one criterion of five here — it's the reason this publication exists, so it carries real weight in the rating, not only in the words. Two rules make that concrete:
- A tool can't earn five stars — ⭐⭐⭐⭐⭐ Excellent — without scoring at least 4 out of 5 on Australian Fit. A globally brilliant tool with no AUD pricing, no local data residency, and no Australian support can still score well on the other four criteria and earn a strong recommendation — but we won't call it Excellent for Australians, because for an Australian it isn't. Where this gate applies we still show the full criteria total; we simply cap the star rating and tell you why.
- A tool can't be a category Top Pick without scoring at least 4 out of 5 on Australian Fit. It can be recommended, even enthusiastically — but the tool we put at the top of a category has to be one that genuinely serves Australians.
So you'll occasionally see a tool with a high total — say 21 out of 25 — wearing four stars rather than five. That's the gate doing its job, not a maths error. The number tells you how good the tool is; the stars tell you how good it is for you, here.
One thing the gate never does: push a tool below its scored band. It only ever removes the fifth star. A tool that scores 16 out of 25 sits in the four-star Very Good band no matter its Australian Fit — even 1 out of 5 — because the gate governs only the top rating and the category Top Pick, never the bands beneath. So when you see a tool with weak Australian Fit at three or four stars, it earned that band on the maths; its local-fit weakness is the reason it can't be crowned the best, not the reason it scored where it did.
We show our working — the score rail on every review
We don't just publish a final number. Every full review carries a score rail. It shows the total, the star band after the Australian Fit gate, and a bar for each of the five criteria. It also carries a "Verified" date and a link straight to the verdict. Beside it sit three facts: what the tool is best for, the AUD price it starts at, and where your data goes.
We leave a fact out rather than guess it. If we can't read a "from" price off the vendor's own pricing page, the rail says so. It points you at the pricing section of the review instead. Quote-only vendors are shown as quote-only. The rail never carries a number we invented.
Under the rail, every review breaks the total into all five criteria with a one-line reason for each mark. You can see exactly where a tool earned its points and where it lost them. Then you can decide whether the criterion that cost it matters for your use. A tool that scores poorly on Ease of Use but well on Feature Depth can suit a power user and fail a beginner.
The breakdown looks like this:
| Criterion | Score | Why |
|---|---|---|
| Value for Money | 3/5 | Fair if you use the full feature set; expensive for light use. |
| Ease of Use | 4/5 | Useful within minutes for most people; a learning curve for new users. |
| Feature Depth | 4/5 | Genuinely deep; one or two gaps against best-in-class. |
| Australian Fit | 2/5 | GST handled, but USD-only pricing, no local data residency, no AU support. |
| Reliability & Support | 4/5 | Strong uptime and status reporting; support is web-only. |
| Total | 17/25 ⭐⭐⭐⭐ |
Australian Fit sits in bold because it's the line we want you to read first.
What the "Verified" date means — and what it doesn't
The date on the rail means one thing. On that day we re-checked the review's claims against the vendor's own pages.
It is not the day the file was last touched. Housekeeping — adding a staleness flag, fixing a typo, restating a price basis — never moves it, because nothing was re-verified. If a review was tidied in September but last checked against the vendor in March, the date says March. A date that flatters us is worse than an old one.
Update history — a price change never erases the old price
Where a review has been updated since it was written, it carries a dated Update history at the foot of the page. When a price, plan, tier or billing currency changes, we add a dated bullet. It gives the old figure and the new one, with the vendor source. The pricing table shows what the tool costs today; the bullet keeps what it cost before.
That old-figure rule applies to changes from September 2026 onwards. Earlier updates are dated and described, but not all of them record the previous price.
The first-reviewed date never moves. It records when we first looked at the tool, and no later update rewrites it.
What the price on a review actually means
Where a vendor bills Australians in Australian dollars, we publish that price as the vendor states it. That includes a vendor's own "inc GST" wording — it's their claim, not ours.
Where a vendor bills only in US dollars, we convert at the exchange rate on the day we wrote the review. We cite that rate and its date on the page. Those converted figures are shown ex-GST, because a GST-registered Australian business that supplies its ABN isn't charged GST on an imported digital service. If you're not GST-registered, allow roughly 10% on top.
Exchange rates move. A converted figure is a guide to what you'll pay, not a quote. The vendor's checkout is the quote.
Hands-on or desk-researched — and why we say which
Of the tools we've scored, fewer than one in five clear the Australian Fit gate at 4 out of 5 or better. That gap is the reason this publication exists.
We can't sit down with 668 tools, and we're not going to pretend we have. Most reviews here are desk-researched. They are built from the vendor's own pricing pages, documentation, help centre and release notes, plus independent reporting. Some are hands-on — usually the tools we use ourselves to run this publication.
Both approaches can produce an honest review. Only one of them can tell you how a tool feels on day three. So you deserve to know which one you're reading.
We are adding a plain line to every review stating its basis: trialled, or desk-researched from primary sources. Most reviews don't carry that line yet. Where a review states its basis, take it at face value. Where it doesn't, assume desk research unless the review describes hands-on use.
How we source and verify
Editorial credibility is the publication. The day a reader catches us with a fabricated price, a wrong plan name, or an unverified feature claim is the day we lose the only thing that separates us from the dozen AI tool aggregators already on the internet.
So every factual claim in every review is anchored to a primary source — meaning the vendor's own pricing page, blog, help centre, or press release. Direct quotes carry a URL and a date. Plan names match what the vendor calls them on their own site. Pricing in AUD that doesn't come from a vendor's AU-specific page is flagged as an estimate — converted at the current USD→AUD rate, cited with its date, and shown ex-GST, because a GST-registered Australian business that supplies its ABN is not charged GST on an imported digital service (buyers who aren't registered should allow about 10% on top) — not presented as a published price.
We use established AU and global media — AFR, ITNews, ARN, iTWire, SmartCompany, InnovationAus, ZDNet AU, The Verge, Wired, Reuters, Financial Times, Bloomberg — for context and cross-checks. We don't use programmatic SEO sites or affiliate-aggregator domains as verification, regardless of how high they rank.
Every primary source we cite is captured to the Wayback Machine on the day of the review. If a vendor later changes a price or a feature claim, we have a defensible record of what the page said when we wrote about it.
The "last reviewed" rule
AI tools change constantly. Pricing, plan tiers, model versions, AU availability — any of these can shift in a quarter. Every review on the directory carries a "last reviewed" date at the top. If a review is more than six months old, it gets a staleness flag until it's been re-verified by us against the vendor's current pages.
We don't quietly update old reviews to make them look fresher. When we change a score or a verdict because the tool itself has changed, we say what changed and when. When we get something wrong, we publish a correction, not a silent edit.
What we don't do
A few things we explicitly don't do, so you know what kind of publication this is.
We don't accept payment for positive reviews. No vendor has ever paid to be included on the directory or to influence a score. If we ever do paid placements, they will be clearly labelled as sponsored and walled off from editorial. The reader's trust is worth more than any sponsorship deal.
We don't run affiliate links covertly. Where an affiliate link exists, it is disclosed inline on the page that carries it and listed on our Affiliate Disclosure page. As at the date at the foot of this page we hold no active affiliate relationships, so no link here earns us a commission. Affiliate revenue will never influence which tools we review or how we score them.
We don't repackage press releases. When OpenAI, Microsoft, or Google launches something, we don't publish three hours later with a recap. We trial a tool hands-on where we can. Where we can't, the verdict stands on the vendor's own documentation and independent reporting, not the press release. Either way, the view is our own.
We don't write generic AI cheerleading. Every review ends with a verdict — including the ones we're not enthusiastic about. If we can't find anything genuinely wrong with a tool, we haven't looked hard enough.
Categories of review you'll see on the directory
We publish four kinds of editorial content. They're built differently and serve different reader needs:
- Full reviews (1,500–2,500 words) — single-tool deep dives with the full five-criteria score.
- Category roundups (longer) — "best of" pages for a category, with quick-pick tables, mini-reviews, and a top pick + runner-up.
- Comparisons (~1,500 words) — head-to-head between two tools that real Australian readers are choosing between.
- News and updates (300–400 words) — short posts on material changes to tools we cover (pricing, AU availability, major features).
The scoring framework above applies to full reviews. Roundups and comparisons rely on the existing scores rather than re-scoring tools mid-piece.
A note on Legal, Finance, and HR tools
Tools in these categories carry an extra duty of care. We will never write about a Legal, Finance, or HR AI tool's accuracy, safety, or compliance without verifying claims against primary sources, and every review in these categories carries a clear note that the tool does not replace qualified professional advice. AI-generated legal interpretations or financial analysis should always be reviewed by an actual lawyer, accountant, or professional.
Automated decisions and Australian Fit
Some AI tools don't just assist a person — they make or heavily shape a decision about one: who gets shortlisted for a role, what a customer is quoted, whether a claim is flagged for review. Australian privacy law is moving toward requiring organisations to be open about this kind of automated decision-making, and we think readers deserve the same transparency from the tools themselves.
So where a tool makes or substantially assists decisions about people — most often in Legal, Finance, and HR — we treat how openly the vendor explains its automated decision-making as an Australian Fit question, and we say what we find in the review.
Corrections, updates, and tools we should review
If you spot something wrong in a review, we want to know. If a price has changed, a feature has been deprecated, or a tool you rely on is missing from the directory, tell us.
- Corrections: adam@theaidirectory.com.au with the URL of the review and the specific claim that's wrong.
- Tool suggestions: same email, with a short note on what the tool does and why it's worth covering.
- General feedback: also fine — we read everything.
Corrections that meet our standard get acknowledged on the affected review with the date of the change and what was updated. Our Editorial and Corrections Policy sets out the full process, and Complaints Handling covers what to do if our answer doesn't satisfy you.
Who writes this
The AI Directory Australia is edited by Adam. Every review, roundup, comparison, and news post is bylined by Adam — there is no agency, no syndication, and no ghost-written content from vendors. AI tools assist with research, sourcing, and drafting; every verdict, score, and final edit is human and bylined by Adam. The publication has one editor, and the byline is always Adam — he stands behind every verdict that runs under his name.
Frequently asked questions
How does The AI Directory score AI tools?
Every full review scores five criteria out of five — Value for Money, Ease of Use, Feature Depth, Australian Fit, and Reliability & Support — for 25 total. Roundups and comparisons rely on those existing scores rather than re-scoring tools mid-piece.
Does The AI Directory take payment for reviews?
No. No vendor has ever paid to be included or to influence a score. If we ever run sponsored content, it's clearly labelled and walled off from editorial. Reader trust is worth more than any sponsorship deal.
Does it use affiliate links?
Not today. We hold no active affiliate relationships, so no link on this site earns us a commission. Where an affiliate link exists, it is disclosed inline on the page that carries it and listed on our Affiliate Disclosure page. Affiliate revenue never influences which tools we review or how we score them.
Are the reviews hands-on?
Some are. Most are desk-researched from the vendor's own pricing pages, documentation and release notes, plus independent reporting. We're adding a line to every review stating which it is, and most reviews don't carry that line yet.
How are Legal, Finance, and HR tools handled?
These carry an extra duty of care. We verify any accuracy, safety, or compliance claim against primary sources, and every review in these categories notes that the tool does not replace qualified professional advice.
Who writes the reviews?
Adam, the sole editor. Every piece is bylined by Adam. AI tools assist with research and drafting; every verdict, score, and final edit is human.
Last reviewed: 20 September 2026. Methodology is reviewed and updated quarterly. Next review due: 20 December 2026.
Update history
- 20 Sep 2026 — Australian Fit note added on age assurance at the point of install, following eSafety's app-store age checks becoming enforceable on 9 September 2026. No change to the five criteria, the 25-point maths or the Australian Fit gate. Our news post
- 20 Sep 2026 — Quarterly review completed. Affiliate wording made conditional so this page and the Affiliate Disclosure page agree: we hold no active affiliate relationships today. Added the score rail, the "Verified" date rule, the Update-history rule and what a price on a review means. Added a section on hands-on versus desk-researched reviews.
- 7 Oct 2026 — The review count was removed from the Australian Fit paragraph; the share that clears the gate is stated instead. No change to the framework.
- 13 Jun 2026 — Quarterly review completed. No change to the framework.
- 2 Jun 2026 — Australian Fit gate documented in full, including the rule that the gate never pushes a tool below its scored band.
- 14 May 2026 — First published.