What 1,628 citations taught me about AI in marketing
Most AI-marketing advice is written by people guessing. This is what the peer-reviewed research actually shows — and the seven findings business owners should be acting on this quarter.
I started writing peer-reviewed research on machine learning in marketing in 2019. The papers have been cited 1,628 times since, mostly by other academics, but increasingly by industry teams trying to figure out what's actually true about AI marketing versus what's marketing fluff.
This piece is the practitioner's version. Seven findings from the research that have held up across the years, the case studies, and the citations — and what each one means for an Australian business owner trying to make decisions in 2026. None of this comes from a vendor. None of it requires a subscription. Most of it contradicts the AI-marketing advice you've been reading on LinkedIn this year.
1. AI in marketing is adopted in three waves, not all at once
In a 2019 paper that has been cited 293 times, I traced the adoption curve of machine-learning analytics in digital marketing across surveyed organisations. The pattern wasn't a single jump. It was three distinct waves: first, automation of reporting; second, augmentation of decision-making; third, autonomous campaign management.
Most AU SMBs are still in wave one. They've automated their reports. They have not yet moved to AI-augmented decision-making — letting AI suggest what to do next, then approving it. That's the current frontier. Skip wave two and you'll skip the period when AI delivers the highest ROI per dollar of effort. The businesses that go straight from automated reporting to "let the AI run the campaign" lose money. The businesses that take the time to learn what good AI suggestions look like, before approving them, build durable competitive advantage.
2. Big data without big interpretation is worse than no data
My 2020 IEEE Access literature review (260 citations) examined more than 50 studies on big data in marketing transformation. The most consistent finding across the literature was unexpected: more data, without a corresponding investment in interpretive capability, made marketing decisions worse, not better. Teams drowned in dashboards, mistook correlation for causation, and over-trusted patterns that were actually noise.
The implication for SMBs is unwelcome. Buying another data tool will not fix your marketing. Hiring or contracting a person who can interpret the data you already have, will. AI tools are a force multiplier on interpretive skill — they amplify whatever interpretive capability already exists. They do not create it. If your current marketing decisions are based on gut feel, adding GA4 + Looker + an AI insights tool will produce gut-feel decisions dressed up in a chart.
3. Chatbots fail when they replace humans, succeed when they triage them
In a 2021 paper on chatbots in digital business transformation (148 citations), my co-authors and I reviewed implementations across multiple industries. The pattern was striking: chatbots that tried to replace human service almost always reduced customer satisfaction. Chatbots that triaged inquiries and routed them to humans almost always improved both satisfaction and operational efficiency.
This is directly applicable to AI marketing in 2026. The temptation to deploy fully autonomous AI agents — chat, lead qualification, customer service — is high. The research says: don't. Use AI to triage and prepare. Keep a human in the loop for the moment of trust. The 2021 finding has held up against every subsequent generation of chatbot technology, including the LLM-grounded systems we have now. The technology is better. The pattern is the same.
4. The biggest barrier to AI adoption is organisational, not technical
Across multiple papers — including my 2022 work on digital business transformation (34 citations) and my 2023 paper on barriers to university-industry collaboration (35 citations) — the same finding recurs: the technical barriers to AI adoption are smaller than the organisational ones.
The companies that succeed with AI marketing are not the ones with the best tech. They are the ones whose leadership tolerates the messy 6-month learning period during which the AI gets things wrong, the team complains, and the early experiments don't pay back. If you can't afford to be patient with AI, you can't afford to do AI. Hiring a more expensive vendor doesn't shorten the learning period; it just compresses your runway.
5. Search visibility frameworks need to be updated for AI search
My 2019 paper on optimisation models for organisational search visibility was written before AI Overviews existed. Reading it back in 2026, almost every framework still applies — the entities, the relationships, the authority signals — but the surface they apply to has changed. Optimising for keywords is now optimising for entities. Optimising for backlinks is now optimising for cited authority. The principles travel; the tactics don't.
Practical implication: if your SEO agency is still selling you keyword-density reports and backlink-volume targets, they're applying 2018 tactics to a 2026 substrate. The substrate now is what AI engines cite, not what Google's classical ranking algorithm preferred. The seven-year-old principles still work — be cited, be authoritative, be the entity people refer to — but the tactical implementation has changed almost entirely. We covered the practical version of this shift in our piece on AI Overviews and traffic loss.
6. Trust signals matter more in B2B than B2C, and that gap is widening
In our 2019 research on knowledge management in Australian companies (Knowledge Management Research & Practice, 24 citations), my co-authors and I found that B2B buyers placed roughly 2.4x more weight on author authority and institutional affiliation than B2C buyers did. Every subsequent paper has reinforced this.
In an AI-search world, where any answer can be auto-generated, that gap will widen further. If you sell B2B, your named expert authors are now the single largest competitive moat you can build. The reason is mechanical: when an AI engine has to choose between a generic, anonymous "best practices" article and a piece with a credentialed expert author and verifiable third-party citations, it picks the second one. Every time. The gap between "anonymous content" and "named-expert content" used to be a marginal trust signal. It's now a binary inclusion criterion for AI summaries.
7. Sustainability disclosure is becoming a marketing channel
In our 2021 papers on environmental sustainability disclosures by ASX-listed companies (44 citations and 35 citations), we observed that the companies most willing to publicly disclose sustainability data were also the ones gaining disproportionate trust signal in B2B procurement. Five years on, this pattern has accelerated.
AI search engines weight disclosed, structured, verifiable claims more heavily than implied claims. Hidden virtue does not rank. If your business has practices worth disclosing — environmental, social, operational — publishing them in structured form on your site improves AI-search visibility as a side effect of doing the right thing publicly. Most AU businesses we audit are doing the work but not disclosing it. That's free moat being thrown away.
The pattern across these findings, accumulated across 15+ years of research and 1,628 cumulative citations, is consistent. AI in marketing rewards organisations with three traits: patience with messy learning periods, investment in interpretive capability over tool acquisition, and willingness to make their expertise legible — through named authors, structured data, and public disclosure. None of these are technical problems. All of them are organisational ones. The agencies and SMBs that succeed with AI in 2026 will be the ones who already had these traits; the AI tools just made the gap visible.
What to do this quarter
- Audit your AI maturity wave honestly. Most AU SMBs are stuck in wave one. Move to wave two: AI-augmented decision support. Pick one decision (campaign budget reallocation, content topic selection, ad creative variant testing) and let AI suggest, you approve.
- Stop buying tools. Hire interpretive capability — internal or external. One good analyst beats three more dashboards.
- Triage with AI; convert with humans. Map every customer touchpoint and identify which are triage and which are trust. Apply AI to the first; protect the second.
- Build named-author authority on your website. If you have an expert internally, surface them. If you don't, find one to partner with. Anonymous content is a 2018 strategy.
- Publish what you know. Disclosed authority outperforms hidden authority in every dataset I've studied.
Want to know which wave you're in?
Catnap audits AU businesses against the three-wave AI marketing maturity model. Honest assessment, prioritised next-move list, no upsell to a 12-month retainer. Most audits take 90 minutes of your time.
The bottom line
AI in marketing is not new. The peer-reviewed literature has been mapping it since at least 2015. The patterns are reasonably stable. The current AI-tool boom is mostly the operationalisation of what was already known. If you ground your decisions in the patterns the research has already found, rather than the headlines this week, you'll spend less, learn faster, and get further than the businesses chasing the trend.
About the author
Dr Andrej Miklosik is the founder of Catnapweb and a Professor at the Department of Marketing and Commerce, Faculty of Management, Comenius University in Bratislava. His full publication list is on Google Scholar; his Web of Science record is O-7582-2016.
Read more about Andrej →Sources & cited works
- Miklosik, A., Kuchta, M., Evans, N., & Zak, S. (2019). Towards the adoption of machine learning-based analytical tools in digital marketing. IEEE Access, 7, 85705–85718.
- Miklosik, A., & Evans, N. (2020). Impact of big data and machine learning on digital transformation in marketing: A literature review. IEEE Access, 8, 101284–101292.
- Miklosik, A., Evans, N., & Qureshi, A. M. A. (2021). The use of chatbots in digital business transformation: A systematic literature review. IEEE Access, 9, 106530–106539.
- Evans, N., Miklosik, A., Bosua, R., & Qureshi, A. M. A. (2022). Digital business transformation: An experience-based holistic framework. IEEE Access, 10, 121930–121939.
- Miklosik, A., & Evans, N. (2021). Environmental sustainability disclosures in annual reports of mining companies listed on the Australian Stock Exchange (ASX). Heliyon, 7(7).
- Miklosik, A., Evans, N., Hasprova, M., & Lipianska, J. (2019). Reflection of embedded knowledge culture in communications of Australian companies. Knowledge Management Research & Practice, 17(2), 172–181.
- Andrej Miklosik — full publication list: Google Scholar profile.