A four-part SEO taxonomy is a map of topics, not a test of results
The review selects 28 articles and organizes AI applications around four SEO areas; it reports no new citation experiment (Ziakis and Vlachopoulou, 2024).
The chapter’s 28 selected articles support a review of AI applications in SEO. They are not 28 experiments showing that a common intervention improves visibility in AI answers. Ziakis and Vlachopoulou, 2024, §2, printed p. 392, volume PDF p. 395.
The chapter offers a vocabulary for organizing a broad subject. Its categories are useful when you need to distinguish page content, external references, local discovery and spoken queries. The difficulty begins when a map of those subjects is presented as proof that a particular product or publishing tactic works.
What kind of review is this?
The authors describe a systematic literature review guided by PRISMA. They began with Scopus, expanded to grey literature and snowballing, screened titles and abstracts, and settled on an AI-and-SEO review set. The chapter also uses additional references to introduce SEO techniques. [§2, printed p. 392].
That final point matters. A reference list can contain background works as well as included review studies. Counting every numbered reference as an independently tested AI-visibility intervention would misrepresent both the selection process and the kinds of work included.
The chapter covers AI used by search systems and AI used by marketers. Those are related but distinct subjects. A system that improves query interpretation is not the same intervention as a marketer changing a page. A tool that helps draft text is not, simply by existing, evidence that the text receives more citations.
The chapter appeared in a 2024 Springer proceedings volume. Its listed access dates include 28 July 2023 for several web references. Publication year should therefore not be read as a guarantee that its practitioner examples describe the latest search products or policies. [Title page, printed p. 391; references, pp. 398 to 399].
Why do the screening counts change?
| Stage described in the chapter | Reported count | What the count means |
|---|---|---|
| Initial Scopus search | 33 | Articles returned by the foundational search |
| Broadened collection | 73 | Pool after adding grey literature and snowballing |
| Title and abstract screening | 44 | Articles retained after that stage |
| Final AI-and-SEO review set | 28 | Articles forming the stated review backbone |
Source: §2, printed p. 392, volume PDF p. 395. These are selection-stage counts. They are not shares of successful interventions, so a Wilson interval is not applicable.
The broadened pool follows the initial search. It is not described as an independent additional sample to add to the initial count. Nor should the final set be added to the screened set. Each number belongs to a stage, with a different relationship to the material ultimately discussed.
The query is printed as a combination of Artificial Intelligence and the American or British spelling of Search Engine Optimization in title, abstract and keyword fields. The chapter does not provide a fully exported, executable search history with all database settings and a record-level inclusion table. A fresh search today would require recording its own exact syntax and date. [§2].
PRISMA is a reporting framework. Naming it does not establish that every included source is empirical, that the search is exhaustive or that the included claims have all been independently reproduced. Those questions still require the search record and the individual sources.
What are the four areas?
| Area in the chapter | Main subject | Question the category does not answer by itself |
|---|---|---|
| On-page SEO | Content, metadata, technical page features and user experience | Does a particular page change cause more citations? |
| Off-page SEO | Links, references, reputation and external presence | Which external reference changes a measured outcome? |
| Local SEO | Geographic discovery, listings and local reputation | Does this apply to a global B2B software question? |
| Voice search | Spoken, conversational queries and their interpretation | Does a voice-answer finding transfer to a text-based AI answer? |
Source: §§3.1 to 3.4 and Figure 1, printed pp. 393 to 395, volume PDF pp. 396 to 398. The chapter calls these the “four pillars of modern SEO.” The phrase is its organizing frame, not a set of experimentally estimated factors.
Categories also overlap. A mobile-friendly page can matter to a local visitor. A review can be an external reference and a source of local information. Spoken queries can reach content that also serves ordinary typed searches. The taxonomy should help identify the question, not imply that the categories are statistically independent causes.
For a software marketer, that is a useful restraint. You can use the frame to ask where an activity belongs while still requiring separate evidence that the activity changes the outcome you care about.
What evidence would a claim about improved visibility need?
First identify what visibility means. A page can be retrievable, indexed, displayed as a search result, cited as a source or associated with a named product in an answer. The chapter discusses a broad range of search applications without establishing a single common outcome scale across them. [§§3 to 4].
Next identify the comparison. A before-and-after observation has different limits from a controlled experiment. A prediction model evaluated on an existing dataset has different limits again. A practitioner explanation of a tool’s possible use does not supply either comparison merely because it is cited beside a research article.
The chapter does not report a pooled intervention effect, a common sample of B2B buying questions or a new multi-engine source-citation experiment. It is therefore unsuitable as the numerical basis for a promise that using AI in SEO increases citations by a particular percentage. [§§2 to 4].
This does not mean that reviews cannot provide evidence. A well-specified synthesis can assess patterns, disagreements and uncertainty across studies. Here, the correct use is narrower: a narrative overview and classification that points the reader toward claims requiring their own primary-source checks.
Which statements deserve a separate source check?
The on-page section discusses schema, image optimization, semantic methods and content assistance. The off-page section discusses reputation monitoring and sentiment analysis. The local and voice sections discuss geographic patterns and conversational interpretation. Each application needs its own evaluation before being treated as a verified route to an outcome. [§§3.1 to 3.4].
For example, automatically producing markup and producing accurate, appropriate markup are different achievements. Producing valid markup and earning a search feature are different again. The review’s description of automation does not establish that an individual site qualifies for a feature or that a citation system uses that markup as a selection signal.
Similarly, identifying sentiment in reviews is an analytic task. It does not establish that changing sentiment causes a search engine to cite a product. A B2B company can find the analytic task useful without converting it into an untested citation strategy.
The chapter itself identifies the opacity of proprietary search algorithms as a limitation. That supports caution about universal claims. It does not remove the need to inspect whether a specific cited study measured the particular effect being advertised. [§4, printed p. 396, volume PDF p. 399].
How does the review compare with a dated literature census?
Its scope is AI and SEO broadly. Our separate literature census asks a narrower question about empirical research on source selection or citation in AI answers, with its search date, screened records and venue checks disclosed. The two counts cannot be compared as if they measured growth in one unchanged field.
The census also separates candidates from jointly verified works. An abstract that looks relevant is not automatically a fully verified empirical paper. A published venue record, a primary paper and its measurement description answer different parts of that check. The chapter’s selection-stage counts are no substitute for those checks on a later body of literature.
The remaining open question is about effects, not vocabulary. A fictional company, Cedarline Software, could classify its documentation work as on-page activity and its review-profile work as off-page activity. It would still need dated observations or a suitable experiment to assess whether either affected citation or naming under a defined question set.
What our own data shows
In our registered citation study, the ten most frequent exact hosts account for 3,192 of 18,893 target citation occurrences, or 16.9%, descriptive 95% Wilson interval 16.4% to 17.4%. Absence Index release 2026-09; Study 2 methods and results.
That is a distributional observation under the September question set. It does not test the chapter’s four-part taxonomy or identify the effect of any proposed optimization. We have no controlled measurement of a citation gain from adopting the taxonomy on this point.
Sources
- Ziakis, C., & Vlachopoulou, M. (2024). Artificial intelligence’s revolutionary role in search engine optimization. In A. Kavoura et al. (Eds.), Strategic innovative marketing and tourism (pp. 391 to 399). Springer. https://doi.org/10.1007/978-3-031-51038-0_43 . Methods p. 392, taxonomy pp. 393 to 395, discussion p. 396; supplied volume PDF pp. 394 to 402. Verified 30 October 2026.
- Broadcastwell. (2026). Absence Index, release 2026-09 [Data set]. Zenodo. Release JSON, answer JSONL and README. https://doi.org/10.5281/zenodo.22907695 Verified 30 October 2026.
Cite this page
Broadcastwell research team. (2026, October 30). A four-part SEO taxonomy is a map of topics, not a test of results. Broadcastwell. https://docs.broadcastwell.com/research/sources/ziakis-vlachopoulou-2024/ Published 30 October 2026; updated 30 October 2026.
@online{broadcastwell2026research19,
author = {{Broadcastwell research team}},
title = {A four-part SEO taxonomy is a map of topics, not a test of results},
year = {2026},
date = {2026-10-30},
url = {https://docs.broadcastwell.com/research/sources/ziakis-vlachopoulou-2024/},
urldate = {2026-10-30},
note = {Published 2026-10-30; updated 2026-10-30}
}
Prepared by the Broadcastwell research team. Lead: Ethan Brooks, management consultant.
Sources checked 30 October 2026. Independent check: 30 October 2026, Broadcastwell verification desk. This page has not been peer reviewed.
Last updated: 30 October 2026. Next review: 30 January 2027.