Ninety four percent of B2B buyers used AI during their most recent purchase process. Fifty five percent used it to compare vendors, fifty four percent to research products, forty seven percent to build the internal business case. All of it happened before any vendor was contacted. In parallel, fifty one percent of B2B tech brands have zero citations across ChatGPT, Perplexity and Gemini. Nearly three quarters of ChatGPT's B2B vendor citations come from earned media, not owned content.
Read those numbers in the right order and the picture snaps into focus. The Australian buyer for enterprise software, cybersecurity, cloud, data or AI services is now assembling the shortlist inside an interface the vendor cannot see, cannot instrument, and in most cases is not present in. The deal is decided before the sales team knows there is a deal.
This is not a marketing problem. It is a pipeline problem that has been hiding inside dashboards that measure the wrong thing.
What actually changed in the buyer journey
Google Search referral traffic to publishers has been in a measurable slide for eighteen months. Ahrefs' February 2026 study, run across 300,000 keywords with December 2023 and December 2025 as the endpoints, found that AI Overviews correlate with a 58% reduction in click through rate for the top ranking page. When an AI Overview sits at the top of the results, only 1% of users click a link inside it. Only 8% click a blue link below it, compared to 15% when no AI Overview is present. AI Overviews now trigger on roughly 48% of tracked search queries. Sixty percent of all Google searches end without a click. For queries where an AI Overview appears, the zero click rate hits 80 to 83%.
Publishers moved first, because they measure traffic in real time. Digital Content Next member sites reported median year on year Google Search referral traffic down 10% over an eight week window in mid 2025. Non news brands took a 14% hit. Global publisher traffic from Google was down 33% in the year to November 2025. US publishers took a 38% hit. Antitrust filings in the United States now cite a 58% publisher click decline.
Australian B2B tech companies are inside that same funnel. They just have less visibility into it, because they measure their content in demo bookings and pipeline attribution rather than reader sessions. The Google Analytics numbers look softer than they did in 2024, but the story the CMO tells the board still runs on the assumption that if the site ranks, the traffic follows. It does not.
The shortlist is now assembled inside the model
The more interesting shift is not what Google is doing. It is what buyers are doing in ChatGPT, Perplexity, Gemini and enterprise copilots.
ChatGPT crossed 900 million weekly users in early 2026. Buyers of enterprise software category by category now open a chat window, type "best privileged access management for a mid sized Australian bank" or "how does Wai compare to competitors for AI visibility", and receive a named recommendation set. Perplexity, the most citation transparent of the engines, lists every source it drew from. Gemini fuses its own index with the model's training data. Enterprise copilots layered on Microsoft 365 or Google Workspace pull from a mix of licensed content, first party data and the wider web.
The shortlist that comes out of that interaction is the shortlist that reaches procurement. Seventy two percent of B2B software buyers now use ChatGPT specifically to evaluate vendors. In many categories, the vendors that make the AI generated recommendation are the only vendors that get evaluated at all. The rest never appear in the RFP.
This is a different game to search visibility. The AI engines are pulling from a narrower band of authoritative content, weighted toward earned media, trade press, analyst commentary, structured technical documentation, and clearly attributed expert answers. Marketing pages, product tour videos and gated white papers barely appear in the citation trail.
Why the current SEO stack does not close this
The default response most Australian B2B marketing teams reach for is to publish more, faster, targeting more long tail keywords. That fixes the wrong problem. The bottleneck is not volume, it is citation quality.
AI answer engines pick sources based on a stack of signals that overlaps with, but is not the same as, classic SEO. They favour clarity of assertion. They favour distinct, quotable statements. They favour cross corroboration across independently authoritative sources. They favour entity association: the vendor's name appearing alongside the category, the customer, the regulator, the technology stack. They favour structured content that can be lifted as a chunk without ambiguity.
A well ranked blog post that reads like a marketing brochure will not make the citation set. An article that is genuinely useful, opinionated, quotable and cited elsewhere often will. The playbook that worked in 2019 to reach position one on Google is not the playbook that puts the brand into the ChatGPT answer in 2026.
That is what generative engine optimisation, or GEO, means in practice. It is the discipline of building content, technical structure and citation footprint so that the AI engines have a reason to name the brand when a buyer asks the category question. GEO is not a rewrite of SEO. It sits alongside it, and increasingly ahead of it, because the buyer touch happens earlier.
What Australian B2B buyers actually see
The exercise every Australian CMO, Head of Growth or Head of Product Marketing should run this quarter is the same exercise Wai runs for its ARC clients on day one. Take the ten highest value buying queries in the category. Run each of them through ChatGPT, Perplexity, Gemini and Google AI Overviews. Record every source cited. Record every competitor named. Record whether the client brand appears at all, and if so, in what position and with what framing.
The output is usually uncomfortable. Categories where the vendor holds strong Google rankings frequently return zero citations in the AI engines. Categories where the vendor invested heavily in gated content return citations weighted almost entirely to competitors who published open, indexable, cross linked commentary. Regional buyer queries that carry "Australia", "APRA regulated", "ASX listed" or "APS ready" often surface completely different vendors to the global equivalents, because the AI engines are picking up local authority markers that global brands have neglected.
The engines also disagree with each other. A brand cited by Perplexity is often invisible in ChatGPT. Gemini frequently surfaces Google properties disproportionately. Enterprise copilots add another layer, because they weight first party organisational content ahead of the open web. A vendor that appears in the public ChatGPT answer might still be invisible inside a customer's own Microsoft 365 Copilot response, because the customer never received a piece of collateral the copilot could ingest.
Each of these gaps is a specific, addressable pipeline leak. None of them show up in the metrics the marketing team is currently reporting.
The five inputs that decide whether the engines cite you
Across the AI engines that matter for Australian B2B buyers, five inputs consistently separate the brands that get cited from the brands that do not.
The first is entity clarity. The engines need to associate the brand unambiguously with the category, the geography, the customer segment and the technical stack. That means the brand's name, product names, platform names and category descriptors have to appear together in structured, indexable content. Homepages that rely on video and JavaScript to communicate positioning fail this test. Content hubs with clear taxonomy pass it.
The second is answer quality. The engines favour direct, quotable statements that can be lifted verbatim as answers. Marketing prose that hedges every claim underperforms. Sharp, specific, defensible assertions outperform. The best performing pieces read like commentary from a senior operator, not a corporate voice.
The third is corroboration. Citations flow to brands whose claims are supported across multiple independently authoritative sources. Earned media, analyst mentions, regulatory filings, customer references in trade press, conference talks that get transcribed, podcast appearances that get indexed: all of these are inputs the engines weigh. Vendors that rely only on their own site for the citation footprint underperform, badly.
The fourth is technical extractability. Clean HTML, sensible heading structure, structured data, canonical URLs, unblocked crawlers, fast rendering and clear content chunks all matter. Many Australian B2B sites still block or gate the content the engines would otherwise cite. That is a self inflicted invisibility.
The fifth is Australian specificity. The engines respond to evidence that a brand is a genuine local operator: references to Australian regulators (APRA, ASIC, OAIC, ACSC), Australian standards (Essential Eight, IRAP, VSA6), Australian customers, Australian offices, Australian case studies. Global vendors with no local footprint frequently lose to local operators in Australian buyer queries, because the local operator is the safer, cheaper, more relevant citation.
The measurement problem
The reason this whole movement has stayed off most Australian boardroom agendas is that the current measurement stack does not surface it.
Google Analytics shows lower traffic, but attributes it to seasonality, algorithm changes or channel shifts. It cannot show a query that ended in an AI answer with no click. Google Search Console shows impressions and clicks for Google's own surfaces, but not for the queries buyers asked in ChatGPT or Perplexity. Standard SEO tools show ranking positions, but ranking one on a query where 80% of users never click is a hollow win.
The measurement stack that reflects the new reality has three additional layers. First, AI Overview presence and citation position for the priority query set, tracked over time. Ahrefs, Semrush, and dedicated GEO monitoring tools now expose this data. Second, cited source frequency inside ChatGPT, Perplexity and Gemini for the same queries, sampled at cadence and diffed for movement. Third, a share of voice metric that reports what proportion of buyer relevant AI answers name the brand, and in what position relative to competitors.
Once these three layers exist, the board can see the pipeline exposure. Not the vanity traffic loss, the actual competitive position inside the interface where the buyer is deciding.
What the next 90 days should look like
For an Australian B2B technology brand that has not run this exercise, the first ninety days should not be spent commissioning a rebrand or a content refresh. It should be spent on four things.
Run the citation audit. Ten to twenty priority buyer queries, four AI engines, current state captured, competitors mapped. This is the baseline.
Fix the technical floor. Unblock crawlers, retire the pieces that gate real content, restructure the highest performing pages so the useful assertions are extractable, add structured data where it is missing, cull the pages that dilute the entity signal.
Publish for citation, not for traffic. Original commentary, opinionated analysis, specific data, Australian context, named customers, named regulators, named technologies. The pieces that read like senior operator commentary get cited. The pieces that read like campaigns do not.
Instrument the measurement. AI engine visibility tracking as a standing report, on the same cadence and to the same audience as the existing SEO and paid dashboards. When the CFO asks about pipeline exposure, the answer should already be on the slide.
Wai builds ARC as the infrastructure layer for exactly this work. ARC is Wai's platform for making organisations discoverable, defensible and consistently cited inside AI answer engines. The Australian B2B brands that treat AI visibility as an infrastructure discipline in 2026, rather than a content campaign in 2027, are the brands that will still be on the shortlist when the buyer opens their next chat window.
FAQ
How do B2B buyers use ChatGPT to research vendors? Buyers open a chat and ask a category question, often with region or use case attached. The engine returns a small set of named vendors with a short rationale for each. Buyers then follow up with comparison questions, feature questions and pricing questions inside the same conversation. Most of this happens before any vendor site visit.
What percentage of B2B buyers use AI to shortlist vendors? Recent research puts it at 94% during the most recent purchase process, with 72% of B2B software buyers specifically using ChatGPT to evaluate vendors. The numbers are trending up quarter on quarter.
How do you get your company cited in ChatGPT and Perplexity? Publish content that is unambiguous, quotable, technically extractable and cross corroborated by earned media and trade press. Make sure entity associations between the brand, category, region and stack are consistent across the web. Do not gate the content the engines would otherwise cite. Perplexity in particular is transparent about its citations, so it is the easiest engine to reverse engineer.
Why is my website traffic dropping since Google AI Overviews launched? Google AI Overviews now trigger on roughly 48% of tracked queries and cut click through rate for the top ranking page by 58%. Zero click rates hit 80 to 83% on queries that show an AI Overview. Publishers report median traffic declines of 10 to 38% depending on category and geography.
How do you measure traffic lost to AI Overviews? Compare Google Search Console impressions to clicks over a two year window on queries known to trigger AI Overviews. Then use dedicated tools that expose AI Overview presence for the priority query set. A missing click on a high impression, AI Overview triggered query is a leak that classic analytics will not label as such.
What is generative engine optimisation and does it work in Australia? GEO is the discipline of engineering content, technical structure and citation footprint so that AI answer engines cite the brand when buyers ask category questions. It works in Australia and often works better than in global markets, because Australian buyer queries carry local signals (regulators, standards, geography) that the engines can use to surface local operators over global incumbents.
How do you appear in AI answer engines as a B2B SaaS brand? Focus on entity clarity, answer quality, cross source corroboration, technical extractability and Australian specificity. Treat AI visibility as a standing infrastructure discipline, not a campaign. Measure it on the same cadence as SEO and paid.
How much has Google search traffic fallen for publishers in 2026? Global publisher traffic from Google is down roughly 33% year on year, with US publishers down 38%. Digital Content Next member sites reported median declines of 10% over eight week windows, with non news brands taking a 14% hit. B2B tech content is inside the same funnel and is showing the same pattern.