Australia's biggest retailers are nearly invisible to AI search

Search is quietly splitting in two

There is the Google you know, and there is the fast-growing slice of buyers who now ask ChatGPT, Perplexity, Copilot or Google's AI Overviews "what's the best X" and never click a blue link at all. Those AI engines read a website very differently from a normal browser, and most sites have done nothing to be readable by them.

So on 23 July 2026 I ran a passive check on the homepages of eight of Australia's best-known retailers - Bunnings, JB Hi-Fi, Chemist Warehouse, Myer, Kmart, THE ICONIC, Kogan and Harvey Norman. Household names, big budgets, mature SEO. I looked at the specific signals an AI engine uses to understand and cite a business. Here is what came back.

What I checked, and what I did not

To keep this honest, the boundaries matter. I checked each retailer's homepage only, on one date, using passive public requests - the homepage itself, its /llms.txt, and its /robots.txt. No login, no crawl of deeper pages, no active probing. These are the same publicly observable signals AuditHQ's GEO and AI-readiness suites read.

I did not ask ChatGPT or Perplexity about these brands and watch them fail to appear. I measured whether the foundations for AI visibility are in place, not whether each brand is cited today. That distinction is the whole point of the caveat further down.

The data

The pattern is consistent across all eight:

• Not one serves an llms.txt file. This is the emerging standard that tells AI models what your site is, what to cite, and what they are allowed to use. Zero of eight.
• Not one declares a position on AI crawlers in robots.txt - no explicit stance on whether ChatGPT's or Perplexity's crawler should read them.
• Most serve no structured data in the HTML an AI crawler reads. Bunnings and THE ICONIC render none at all, confirmed in a full browser. Structured data is the clearest way to tell an AI "this is a business, here is its name and products" - and most of these sites make the AI guess.
• Kogan blocked the request outright. If a plain request hits a wall, there is a real chance AI answer engines see the same wall.

None of these sites are badly built. Their HTTPS and security were fine across the board. The gap is not competence - it is that "can an AI engine read and cite you" is a brand-new audit layer that did not exist two years ago, and almost nobody has run it yet.

The caveat that keeps this honest

Here is what I am not claiming: that if you ask ChatGPT for the best hardware store in Australia today, Bunnings is absent. A brand that large may well be named, because AI engines lean on authority and third-party coverage, and Bunnings has decades of both. The missing llms.txt and structured data do not make a household name vanish overnight.

The narrower, fairer point is this: whatever AI visibility these brands have, they have built none of it deliberately. It is on loan from their existing authority. The moment a competitor invests in being genuinely machine-readable - clear structured data, an llms.txt, content written to be extracted - they start to earn citations the incumbents are leaving on the table. That is a slow-moving risk for a giant. For everyone smaller, it is the whole game.

Why this should worry anyone with clients

If the biggest retailers in the country - the ones with the budgets and the SEO teams - have not touched this layer, every SMB and mid-market client you look after almost certainly has not either. That is not a criticism. It is an open lane. The agencies and consultants who start advising clients on AI-search readiness this year are going to look prescient in eighteen months, because they will have earned citations while everyone else was still optimising for ten blue links.

The 6 signals you can check yourself

You can run this on any site, including a client's, in a few minutes:

• llms.txt present and returning 200 at /llms.txt
• Structured data (JSON-LD) in the homepage HTML, ideally server-rendered so a non-JavaScript crawler can read it
• Content the AI can read without executing JavaScript - view source, not just the rendered page
• A robots.txt stance on AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended)
• Security and trust headers (HTTPS, HSTS) - table stakes, and these retailers all passed
• Clean, non-truncated title and description so both search and AI have a clear summary

FAQ