GEO and AEO for publishers: Matching books to reader needs

For publishers working on generative and answer engine optimization (GEO and AEO), knowing that ChatGPT can describe your book is only a starting point. The more useful question is whether it introduces the book when someone describes what they want to understand, do, or read, without already knowing the title.

We investigated this using The Dementia Caregiver by Marc E. Agronin. Our study combined reader reviews, Kadaxis’s audience and marketing models, evidence of customer demand, public book information, and 104 ChatGPT answers to 47 questions.¹

ChatGPT often recommended other books when asked for caregiving guidance. A question that explained what the user wanted to learn produced a more consistent recommendation of Agronin’s book:

What books explain the main types of dementia in plain language and spell out what each means for family caregiving?

We submitted this same question to ChatGPT three times, as separate requests. Repeating the question let us check whether the recommendation recurred or appeared only once.

ChatGPT named The Dementia Caregiver in all three answers, placing it first twice and second once. Each answer linked to the book’s publisher page. The question had not supplied the title or author.¹

The user was asking in ordinary language for an explanation of the different conditions and their implications for family care. ChatGPT’s recommendations pointed to the book’s chapters on those conditions and on caregiving at different stages.

By contrast, questions asking ChatGPT for a very short first-week guide or portable daily-care checklists produced other books. More detail did not automatically make Agronin appear. What mattered was whether the requested kind of help matched a reason to choose this particular book.

The strongest finding was that a plainly expressed reader need repeatedly brought Agronin’s book into ChatGPT’s recommendations, with support from information already available about it.

The reader’s purpose matters as much as the subject

In our previous article, How books show up in ChatGPT and AI search, I explained how a user’s conversation and information found through search contribute to an answer. For book discovery, that means connecting what the person wants with what ChatGPT can establish about a title.²

A book’s subject category gives only part of that connection. Someone asking about dementia might want to understand a relative’s diagnosis, find questions to take to an appointment, manage an immediate difficulty, or choose a reference to consult over time. Books on the same subject can serve those purposes differently.

In one answer to the plain-language question above, ChatGPT paired Agronin’s book with The 36-Hour Day, the family-caregiving guide by Nancy L. Mace and Peter V. Rabins. It presented Agronin as a way to understand different diagnoses and The 36-Hour Day as a guide to everyday care. Both titles appeared in all three answers to that question.¹

Your book does not necessarily have to displace a familiar title. It may be recommended alongside it because the two are presented as helping with different parts of the reader’s situation.

This applies beyond nonfiction advice. A historical novel’s period may establish its broad relevance, while its atmosphere, pace, treatment of violence, or family relationships explain whether it offers the experience someone wants. A business book may help a manager understand an organization without being an implementation manual. Making those distinctions clear on public pages can help a prospective reader, and the search tools helping them, recognize the fit before they know the title.

Reviews help explain why someone would choose the book

Our review-analysis model identified 17 themes across 24 reviews of Agronin’s book. Several helped explain what readers valued and which expectations the public description should address.¹

A review can support more than one theme.

The last two rows are worth considering together. Eight reviews recognized practical content, while three wanted more actionable detail. Those comments suggest that readers were not all expecting the same kind of help.

An explanation of what is happening can be valuable without being a short procedure to follow during a difficult moment. A reference organized around conditions and stages can be useful without providing removable checklists. The repeated comments about explanation and reference value give us a concrete reason to investigate those uses of the book.

I would retain both the praise and the criticism when deciding what to emphasize. A sample can help a prospective reader judge how practical the material is for their situation. A stronger claim in the description cannot resolve a mismatch between what they expect and what they will find.

The same applies to fiction. Reviews may distinguish readers who enjoyed a novel’s reflective pace from those who expected a fast thriller. That difference can help you describe the reading experience more accurately and investigate the conversations in which the novel belongs.

Audience intelligence connects reader experience with discovery

Kadaxis’s audience-intelligence model uses the review evidence and information about the book to identify potential readers, their situations, and the kinds of help they may seek. Our reviewer-backed marketing analysis compares those needs with how the book is presented. Marketing Bridge develops possible positioning and ways to reach those readers.¹

For Agronin’s book, this work identified uses including getting started after a diagnosis, everyday caregiving, talking with doctors, finding support, understanding dementia, and looking after the caregiver. Each gives us a different reason someone might seek a book, and therefore a different conversation to investigate in ChatGPT.

For your list, the value is in moving beyond a broad category such as dementia, management, or historical fiction. Reviews can reveal a benefit readers recognize that is barely mentioned in the description. They can also reveal when the marketing suggests an experience different from the one readers report. The models turn those observations into specific reader needs to check against the book, customer demand, and the recommendations people receive.

Customer demand helps decide where to focus

We assess customer demand using Kadaxis’s book-market model, which draws on relevant retail results and ratings activity, alongside search-interest data. The search data helps us assess interest in related topics; the book-market evidence helps establish whether relevant books attract reader activity.¹

This matters because a carefully constructed question can make a book appear without showing that many people want the kind of help it describes. We need a reason to pursue the audience, not just a successful prompt.

Understanding different dementia types had strong support in our assessment. Portable checklists are a narrower requirement than general caregiving advice. Interest in the broader subject gives us a reason to investigate that use, but a substantial investment in new checklist material would need a more specific case.

That affects the scale of the recommendation. Making a well-supported benefit clearer on an existing page is a smaller decision than commissioning a new companion product. Demand helps you decide where the work is justified.

Different requests brought different books into the answers

The questions below show the distinction between understanding a subject and needing a particular form of practical help. Each was submitted to ChatGPT three times, without naming Agronin or his book.¹

For the checklist question, ChatGPT named Dementia Home Care and Dementia Caregiving for Domestic Helpers in all three answers. It explained those choices by referring to checklists and forms. For the appointment question, two answers explicitly favored free question lists for the immediate task.¹

That is useful feedback for how you describe a book. Information about working with doctors is not necessarily the same product as a list someone can carry into an appointment. A reader looking for the latter needs to know whether the relevant material is a discussion, reusable questions, or a separate worksheet.

The answer is not always to add the requested feature. First establish what the book provides. If it contains suitable material, describe it and show an example. If the need calls for a different product, decide whether a companion is worth creating rather than trying to make the existing description cover every request.

Asking for help is different from asking what to read

One question even mentioned books without asking ChatGPT to name any:

Do caregiver books include short, ready phrases or lines I can say to calm someone who is confused, agitated, or seeing things?

ChatGPT answered yes and supplied examples in all three responses, but named no titles. Some responses then offered to identify specific books. The user had asked whether books contain that kind of material, and ChatGPT answered the feature question rather than giving a reading list.¹

We saw the same distinction when comparing these two questions:

What should I ask the doctor after my parent is diagnosed with dementia?

Are there books that help family members know what to ask doctors after a dementia diagnosis?

The first received suggested questions and supporting information. The second introduced books. Across the 48 information-seeking answers in the study, no book titles were named, although many responses searched the web and included source links.¹

The linked material included health information, caregiver guidance, and printable resources from organizations such as the Alzheimer’s Association and the National Institute on Aging. In the appointment-focused answers, those resources appeared alongside or instead of books.^{1, 5}

This is not a rule that ChatGPT only mentions books when asked. It shows why you need to distinguish an opportunity for your book to be recommended from an opportunity for your author or website to provide a useful answer.

For the first, a product page should explain why someone would choose the title. For the second, an extract or author article needs to provide useful information in its own right, with a clear connection to the book. A resource page may instead help someone find current assistance. These are different ways your publishing can become visible, and they should be evaluated separately.

The successful recommendations pointed to existing information

The three answers to the plain-language question about dementia types linked to Agronin’s publisher page. They referred to the separate chapters on Alzheimer’s disease, vascular dementia, frontotemporal dementia, and dementia with Lewy bodies, followed by caregiving in mild, moderate, and advanced stages. Those sections are present in the published contents.^{1, 3}

The author’s own page also explains that the book helps readers understand changes in a loved one, different forms of cognitive impairment, and care needs. The publisher’s synopsis emphasizes becoming well-informed and caring for both the person with dementia and the caregiver.^{3, 4}

I would preserve that material. The recommendations show that it can already support discovery. The opportunity is to make the connection between particular sections and a reader’s reason for choosing the book easier to recognize.

A contents list establishes what is covered. A short explanation can show why that coverage matters to a particular reader. An excerpt can then demonstrate the level and style of the material.

For a business book, saying that it includes case studies establishes a feature. Explaining which management decisions those cases help the reader understand is more informative. For fiction, a setting alone does not tell someone whether to expect a thriller, a romance, or a reflective family story. Those distinctions belong in the material people encounter while choosing what to read.

What I would change, and where I would publish it

For Agronin’s book, I would prioritize clearer explanations and examples of its existing value. These are proposed publishing changes based on the findings, not changes whose effects we have measured.

1. Make the family-caregiver purpose explicit in the description

Keep the current caregiver positioning, but connect the intended reader with the conditions covered and the way the book is organized. An addition could be:

For family caregivers who want to understand the condition as well as the care involved, The Dementia Caregiver by Marc E. Agronin covers Alzheimer’s disease, vascular dementia, frontotemporal dementia, and dementia with Lewy bodies. Separate chapters address caregiving in mild, moderate, and advanced stages, alongside mood and behavioral changes and the caregiver’s own wellbeing.

The published contents support those elements.³ The wording explains why they belong together for this reader, rather than simply adding a list of medical terms.

Publish the approved paragraph on the main book page and align the descriptions on the author’s site and in public retailer listings.

2. Give prospective readers something they can inspect

The three reviews seeking more actionable detail raise a question that an excerpt can help answer: what does the book’s practical material actually look like?

I would select an authentic passage showing how Agronin explains a condition or caregiving situation. A short introduction should identify the book and author, explain who the section is for, and tell a prospective reader what they can learn from it. A companion guide to the contents could point readers toward the relevant chapters for different concerns.

Publish this material where it has a clear relationship to the book:

For the symptom material, distinguish changes affecting the person with dementia from the caregiver’s own health. The book addresses both, but they correspond to different reader questions.³

This gives you options beyond revising the main description. A publisher-page excerpt demonstrates the book. An author article can explain a related concern. A reading-list editor can assess a specific reason to include it. The material should contribute something useful at each destination, rather than repeat the same general blurb.

3. Make separate decisions about new companion material

Before promoting appointment sheets, ready phrases, or one-page checklists, establish whether the book already provides them in a usable form. If it does, an authorized sample may be enough to show the feature. If not, a new companion resource is a separate editorial and commercial decision.

The same applies to the resources at the back of the book. The directory questions in our study repeatedly produced maintained online services as well as books and guides. Those answers pointed users toward resources such as the Alzheimer’s Association and AARP for organizations, helplines, and local assistance.^{1, 5}

A resource chapter can remain useful without being a continuously updated directory. If you decide to maintain an online companion, publish it in the book’s resources area, identify its scope and review date, and link it to the title.

These decisions should follow the evidence of reader need and the value of the book. An explanatory reference does not have to become a checklist product to deserve a recommendation.

Why we asked the same questions more than once

We repeated selected questions because the same wording could produce different books, a different order, or different supporting pages. The point was to identify recommendations that recurred, rather than choose a favorable screenshot.

For the dementia-types question, Agronin appeared in all three answers, but it was not always first and the other recommendations varied. In the checklist answers, the same two books recurred while the leading choice changed.¹

For your own books, begin with a complete request someone in the intended audience might make. Submit it without naming your title, keep the wording unchanged when repeating it, and compare the books, explanations, and links. Record separately any check where you supply the title or URL, because that tests what ChatGPT can say about a known book rather than whether it introduces the book itself.

The context matters. In this study, a request for short summaries and deeper chapters omitted both dementia and the caregiver audience. It produced books from other fields. That result does not establish that a dementia book is a poor match for a family caregiver; the user had never given ChatGPT that situation.¹

Use the comparisons to identify a specific decision: a benefit that needs explaining, a feature that needs demonstrating, or a request the book does not serve well. After publishing a change, repeat the same questions and check the recommendations and destinations again.

What this means for your catalog

The strongest finding was that a plainly expressed reader need repeatedly brought Agronin’s book into ChatGPT’s recommendations, with support from information already available about it. Other requests led to different books or free resources because the user wanted a different form of help.

That is what Kadaxis’s models and analysis are intended to make actionable. Review analysis identifies valued material and unmet expectations. Audience intelligence identifies potential readers and their needs. Marketing analysis and Marketing Bridge suggest how to communicate the book’s role. Customer-demand evidence helps prioritize the work, and the ChatGPT answers and sources show what those prospective readers currently encounter.

For your catalog, the result should be a short set of publishing decisions: a description to clarify, an excerpt to release, an author article to commission, a reader expectation to address, or a publicity angle supported by the book.

You cannot give a book the established position of The 36-Hour Day by rewriting its description. But the answers show that another title can be recommended alongside it for a particular purpose. The opportunity is to make those genuine reasons for choosing your book visible on the pages and in the conversations where the right readers can find them.

Study note: This ChatGPT-only study includes 104 saved answers to 47 questions, collected from September 12 to 15, 2026. Questions were asked more than once; the four examples in the comparison table were each asked three times. The research combines analysis of 24 reader reviews, audience and marketing models, customer-demand indicators, public book information, and selected source-page inspections. The counts report what appeared in these answers.

Sources and further reading

1. Kadaxis research, September 2026. The review findings, demand assessment, and ChatGPT comparisons are drawn from the study described above.
2. Chris Sim, “How books show up in ChatGPT and AI search.” Our preceding article explains how trained knowledge, search, and the conversation contribute to book discovery.
3. Bloomsbury, “The Dementia Caregiver.” Publisher description, contents, product details, and review excerpts used to check the book’s coverage.
4. Marc Agronin, “The Dementia Caregiver.” The author’s explanation of the book and its intended readers.
5. Examples of sources linked in the saved answers: Alzheimer’s Association resources, National Institute on Aging appointment worksheets, and the AARP Dementia Resource Guide. These illustrate the kinds of supporting material ChatGPT presented alongside or instead of books.