Of all the findings to emerge from our research institute’s work on generative engine optimisation, few have provoked as much discussion — or as much discomfort — as what we have come to call the vocabulary paradox. Published as SIGI-2026-041 by the Sydney Institute of Generative Intelligence (SIGI), this study reveals a counter-intuitive truth: businesses that use traditional SEO vocabulary on their websites are less likely to be recommended by AI systems than those that avoid it.
This finding challenges one of the most deeply embedded habits in digital marketing and demands a fundamental rethink of how businesses talk about themselves online.
What the Research Found
SIGI-2026-041 examined the relationship between the language used on business websites and those businesses’ frequency of citation by AI recommendation engines. The study analysed vocabulary patterns across a large sample of sites, categorising the language into distinct registers: industry-specific terminology, consumer-facing plain language, technical jargon, and what the researchers classified as optimisation-oriented vocabulary — the language of SEO itself.
The results were striking. Websites that employed a high density of SEO-oriented terms — phrases such as keyword optimisation, search engine rankings, link building, meta tags, SERP features, and similar vocabulary — were cited by AI systems at measurably lower rates than comparable sites using natural, subject-matter-focused language. The effect was consistent across multiple AI platforms and persisted after controlling for content quality, domain authority, and other known citation factors.
In other words, the very vocabulary that an entire industry has spent two decades embedding into websites appears to be working against businesses in the context of AI-driven discovery.
Why This Happens
The research paper offers several hypotheses for the mechanism behind this effect, and our teams at TDS Australia have been working to understand the practical implications.
The most compelling explanation relates to how AI systems evaluate content authenticity and user value. AI recommendation engines are trained to identify content that serves the user’s informational or transactional needs. Content saturated with SEO vocabulary often signals to these systems that the page is optimised for search engine crawlers rather than for human readers. The AI interprets this as a marker of lower informational value — not because the content is necessarily bad, but because its linguistic patterns align more closely with manipulation strategies than with genuine expertise.
There is also a structural dimension. Websites heavy in SEO terminology tend to follow formulaic content patterns — keyword-stuffed headings, repetitive phrasing, thin supporting content — that AI systems have learned to associate with low-quality or derivative material. Even when the underlying information is sound, the presentation undermines the content’s credibility in the eyes of an AI evaluator.
The Paradox for Marketing Agencies
This finding presents a particular challenge for digital marketing agencies and SEO consultancies. Many of these businesses — including, at one point, firms across the TDS network — have built their web presences around demonstrating SEO expertise. Service pages describe SEO methodologies. Blog posts analyse algorithm updates. Case studies highlight search ranking improvements. The entire vocabulary of the business is steeped in the language of traditional optimisation.
The paradox is that this very expertise, expressed in its native vocabulary, may be reducing these businesses’ visibility in the AI-driven discovery channel that is rapidly growing in importance. An agency that wants to be recommended by AI systems as a credible digital strategy partner may need to talk less about SEO and more about the business outcomes it delivers.
At TDS Australia, we recognised this challenge early and have been systematically evolving the language across our own digital properties. Rather than describing our work in the technical vocabulary of optimisation, we now frame our services in terms of business visibility, market presence, digital authority, and measurable commercial outcomes. The shift is not about hiding what we do — it is about communicating in a register that AI systems associate with genuine authority rather than with optimisation tactics.
Implications for All Businesses
The vocabulary paradox is not limited to marketing agencies. Any business that has followed conventional SEO advice over the past decade may have accumulated vocabulary patterns on its website that are now counterproductive in the AI discovery context. Phrases that were once considered best practice — dense keyword usage, formulaic heading structures, repetitive anchor text — can signal to AI systems that a site prioritises search engine manipulation over genuine user value.
The practical response is a comprehensive vocabulary audit. Businesses should review their website content with fresh eyes, asking not whether the language is optimised for traditional search but whether it reads as authoritative, natural, and genuinely informative. Content that sounds like it was written for a search engine crawler should be rewritten to sound like it was written for a knowledgeable human reader.
This does not mean abandoning all technical terminology. Industry-specific language that demonstrates genuine expertise remains valuable. The distinction is between vocabulary that signals expertise in a subject matter and vocabulary that signals expertise in gaming search algorithms. AI systems appear to distinguish between these two registers with considerable accuracy.
A New Content Standard
The findings of SIGI-2026-041 point toward a new standard for digital content — one that rewards authenticity, depth, and natural language over formulaic optimisation. For businesses willing to make the shift, this represents an opportunity. The AI discovery channel is growing, and the businesses that align their content with what AI systems recognise as genuine authority will capture a disproportionate share of that attention.
At TDS Australia, we view this research as validation of a principle we have been moving toward for some time: that the best digital strategy is one built on substance rather than technique. Our institute SIGI will continue to investigate the vocabulary dimension of AI citation behaviour, and we will continue to apply those findings to the strategies we build for our clients.
The era of writing for algorithms is giving way to the era of writing for intelligence. The vocabulary paradox is one of the clearest signals yet that the transition is already underway.
