GEO is not a content trick: how to make a creative practice legible to people and AI search

GEO, or generative engine optimization, is not a trick for forcing a creative practice into ChatGPT, Google AI Mode or another answer engine. The durable approach is simpler and more demanding: publish specific, people-first pages that answer a real question, show first-hand proof, make sources and limits visible, and let standard search fundamentals do their work.

For an independent art director or brand practice, that means writing about real decisions — what a branding package includes, when a business needs a visual system, how an identity-led funnel works, or how AI can be used without hiding authorship. It does not mean producing hundreds of thin keyword pages, inventing credentials or adding magic markup. Visibility is not guaranteed by any paper, tool or platform.

What has changed — and what has not

Generative search experiences can gather information from multiple sources and synthesize an answer. Google describes methods such as retrieval-augmented generation and query fan-out: a model may issue related searches, retrieve pages from the Search index and use them as support for a response.[1]

This changes the shape of discovery. A visitor may arrive after asking a larger question than a classic keyword query: “What should a rebrand include for a local business?” or “How can a photographer use AI without losing a personal style?” A single precise article can therefore support several related conversations.

But the fundamentals have not been replaced. Google states that there are no additional requirements, special markup or `llms.txt` file required to appear in AI Overviews or AI Mode. A page still needs to be crawlable, indexed, useful and available in readable text.[2]

What the GEO research actually says

The 2024 paper GEO: Generative Engine Optimization introduced a benchmark for studying visibility in generative-engine responses. In its experiments, strategies using citations, quotations and statistics could improve particular visibility measures, while results varied by domain and engine. The authors report improvements of up to 40% in their benchmark; this is not a promise of traffic, ranking or citations for an individual site.[3]

The practical lesson is not “add more links.” It is to make a page worth using as a source. A clear answer, a well-defined claim, a relevant source and an original example are easier for a reader — and potentially a retrieval system — to understand than a page built from repeated jargon.

A later preprint by Chu and Hou makes the ethical boundary clearer. In experiments about LLM product recommendations, authority-style marketing language, including fabricated evidence claims, could shift recommendations. The study treats this as a manipulation and verification risk, not as a tactic to copy. It was conducted in specific product categories and models, so its figures should not be generalized to creative services. Its useful lesson is narrower: when systems cannot distinguish between options, real, verifiable differences matter more than synthetic authority.[4]

The five parts of a legible creative practice

1. State the actual question

An Insight should begin with the question a founder or creative is already trying to answer. “What does a branding package actually include?” is more useful than a vague article called “Thoughts on Branding.” It gives the visitor a reason to read and gives the page a stable subject.

The answer should arrive early. Not as a clickbait summary, but as a statement that still makes sense when read alone.

2. Describe the decision, not just the discipline

“Branding,” “art direction” and “AI” are broad labels. A valuable page shows the decision beneath them: when a business needs a diagnostic rather than a logo; which applications matter first; why a funnel is part of a brand experience; when a creative should disclose AI-assisted imagery.

This is where an independent practice can be more useful than a large generic content publisher. It can write from the actual thresholds and trade-offs encountered in work.

3. Attach first-hand evidence with a clear scope

Evidence does not need to mean inflated before-and-after metrics. It can mean stating the exact role, the material used, the status of the work and the limit of the claim.

For example, StreetLife Rocks can be described as a visual-development engagement connecting calligraphy, product concepts, images and participation. Selected social and product images are transparently labeled as AI-assisted concept mockups; they are not presented as manufactured products or campaign outcomes. Golden Frame Circle can be described as an identity-led funnel, without inventing conversion results.

These details make the practice more legible than generic AI imagery or agency language because they identify what was actually made.

4. Make sources and limits visible

A source link is not decoration. Use it when making an external claim about a platform, research result, market or technical process. State when a paper is experimental, when a case has no verified performance data and when a recommendation depends on context.

This is especially important around AI. “GEO can guarantee visibility” is not supported. “AI can make a brand more human” is too vague. A more honest claim is conditional: AI can help explore or produce an image system when it follows a clear direction, real source material, authorship and permission rules.

5. Keep the site technically ordinary and editorially specific

The technical work remains basic but necessary: crawlable pages, readable text, accurate titles and descriptions, internal links to relevant work, responsive design and Search Console once the domain is connected. The editorial work is the differentiator: clear names, useful decisions, authored examples and language that does not sound interchangeable.

Google’s current guidance is explicit: avoid manufacturing many pages simply to chase variations of a query, and prioritize non-commodity, people-first content instead.[1]

What this practice will not do for “GEO”

It will not manufacture testimonials, performance numbers, awards, clinical-style authority claims, scarcity or client outcomes. It will not call AI-assisted concept imagery a real product photograph. It will not build a website around prompt tips while hiding the actual visual judgment behind the work.

This is not only an ethical preference. It is a strategic one. When every site repeats the same artificial authority signals, the signal becomes less useful. Specificity — a real project, a defined role, a material origin, a transparent limit — remains harder to copy.

A practical publishing rhythm

The first useful sequence is small: one article about the scope of branding, one about visual mismatch, one about AI art direction or creative practice. Each should connect to a real offer or case study. Publishing one well-researched article a week is stronger than generating a large archive of generic AI news.

For this site, the next questions are already visible:

• What should happen before a rebrand?

• What makes a brand world feel coherent?

• How can photographers build a visual pipeline without losing their style?

• When should a creative use AI in a client project?

Each is a decision a real person can use. That is the standard.

Limits and sources

The GEO and LLM-recommendation papers cited here study evolving systems, selected models and defined experimental settings. They do not establish a universal formula for visibility, client acquisition or the ranking of creative services. Google also states that meeting its requirements does not guarantee crawling, indexing or serving. The reasonable goal is not to manipulate a black box. It is to become a clearer, more useful and more verifiable source when a relevant question is asked.

If your work is strong but difficult to recognise online, start with the underlying message and visual system. A Brand Diagnostic identifies the first decisions worth making before more content, a new logo or a generic AI campaign.

References

[1] Google Search Central, “Optimizing your website for generative AI features on Google Search”

[2] Google Search Central, “AI features and your website”

[3] Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan & Deshpande, “GEO: Generative Engine Optimization” (KDD ’24)

[4] Chu & Hou, “Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems” (2026 preprint)