Why SEO is dying, and what replaces it in 2026.
Keyword rankings still exist. Traffic from them doesn't. We walk through the data from our first 40 SMB audits, and the three things every small business should be doing instead, starting this quarter.
SEO isn't dead, but the SEO funnel most Australian SMBs were built around is. The drop has been steady, not dramatic, and that's exactly why most owners haven't noticed it yet. By the time the quarterly report lands and the line on the Google Analytics chart is undeniably lower, the buyer behaviour underneath has already shifted to somewhere their site doesn't appear.
We've now run AI visibility audits on 40 small businesses across hospitality, professional services, trades and ecommerce. The pattern is consistent enough that we're prepared to say it out loud.
What we're actually seeing.
Across the 40 sites we audited, organic search traffic is down a median 28% year-on-year, while the same sites' branded search volume is up. People are still looking for these businesses; they're just doing the discovery step somewhere else. Almost universally, they're doing it inside an AI assistant.
- ChatGPT and Claude now resolve roughly half of the category-research queries that Google used to absorb (e.g., "best accountant in Newtown for sole traders").
- Perplexity is becoming the default for buyers who want a cited, source-rich answer.
- Gemini, baked into Google's own SERP, is short-circuiting many of the long-tail informational queries that used to drive blog traffic.
The result: the work that brought the click (keyword pages, blog posts, schema, internal linking) still gets indexed, but the click doesn't land. The AI summarises the answer and the buyer never opens the source.
What replaces it.
Three things, in order of return-on-effort:
1. Make your site quotable. Restructure pages so the answer to the most-asked question in your category appears in the first 60 to 80 words, in two-sentence form, with an unambiguous attribution. AI models extract by paragraph, not by page. If the answer isn't sitting cleanly above the fold, you don't get cited.
2. Publish an llms.txt file. It's the AI-era equivalent of robots.txt: a single plain-text file at the root of your site that tells AI crawlers what they're allowed to use, in what context, and how to attribute it. It's a 20-minute change. We've seen citation rate lift on the audit re-test within four weeks of adding it.
3. Track citations, not rankings. SEMrush, Ahrefs and the rest are still measuring the wrong thing for this transition. Run a monthly query panel across ChatGPT, Claude, Gemini and Perplexity for the 20 to 30 prompts your buyers would actually ask. Score yourself on appearance rate and share-of-voice against the three or four competitors AI is currently recommending in your category.
The thing nobody wants to hear.
This isn't a panic. Buyers haven't stopped buying. The path they take to find you has changed shape, and the agencies who built their business on the old shape are mostly still selling the old playbook. If you're being told to spend another quarter writing 1,500-word blog posts targeted at keywords with declining traffic, you're paying for a strategy that's already losing.
Audit where you stand. Pick the highest-leverage three fixes. Ship them. Re-measure in 60 days. That's the work for 2026.