Optimise Content for AI Citations - A Practical Guide from GPT Insight

July 3, 2025

Marketing Two Cents

Table of Contents (Click to show/hide)

10.54%

Monthly
Active User Rate

$5000

Daily
Campaign Budget

60%

Increase
Click-through Rate

15%

Growth
Return on Investment

Customer
Segmentation

Prioritisation of
Limited Resources

Competitive
Responses

Consumer
Change

‍

Allow me to put my own conclusion here - AEO is SEO, SEO is AEO. Whatever your are doing for SEO, you are already optimising for AI search.

‍

Creating content that's appealing not only to humans but also to AI models is becoming increasingly crucial for online visibility. Recently, GPT-Insights published an insightful article that delves into how Claude, a prominent AI model, searches for and cites external information. Unlike many other AI models, Claude provides transparent insight into its citation behaviour, which offers valuable lessons for SEO content creation. However, it's important to remember that while we can draw analogies, the specifics of Claude's citation logic may not fully apply to other major AI models like ChatGPT and Google's Gemini.

‍

For example:

ChatGPT (OpenAI): Doesn't have fixed prompt categories but performs similar web queries, typically 1–3 parallel searches. However, it tends to link sparingly and contextually.
Gemini (Google): Very little is publicly documented, and source citations are rare within Gemini’s chat interface.
Commonalities across all LLMs: AI models prioritise content that clearly and semantically fits the prompt rather than traditional SEO signals. Quotable, clear, and structured content is critical.

‍

To simplify things, I've summarised the insights from the GPT-Insights article into an easy-to-digest table below. This table clearly outlines the types of content Claude is most likely to cite and recommend—giving content creators practical ideas on prioritising their content if attracting AI attention is a goal.

‍

Content Types Preferred by AI Models for Citation and Linking
Claude AI Action	Meaning & Provided Information	Website Content Type (Ideal)	SEO Implications: Do’s & Don’ts
never_search	AI answers internally without searching; no external source is referenced.	Timeless definitions Basic facts or general knowledge	✅ DO: Optimize for human readers and UX. ❌ DON'T: Expect AI citations.
do_not_search_but_offer	AI answers internally but suggests a user-performed search for fresh or dynamic content.	Real-time statistics (stock prices, economic indicators) Interactive tools (calculators, configurators, planners) Regularly updated tables (exchange rates, price comparisons, product databases)	✅ DO: Regularly update content and mark freshness clearly. ✅ DO: Implement schema markup for dynamic data.
single_search	AI conducts one targeted external search for current factual data and quotes from the source.	Breaking news or recent events (sports results, weather updates) Latest financial news (new cash rates, interest rate decisions) Recent market updates (average housing market values, market trends) Individual user-generated content (reviews, testimonials, ratings) Region-specific or niche content not internally available to Claude	✅ DO: Publish timely, structured, and authoritative content. ✅ DO: Ensure concise, clear headlines and summaries for quick extraction. ❌ DON'T: Include fluff or vague details.
research	AI performs multiple external searches, compiles data, and cites multiple sources to provide detailed and well-rounded information.	Comprehensive guides and tutorials Expert-driven analyses with evaluation, context, or problem-solving Original research, comparative studies, detailed databases Content needing editorial expertise and deep context understanding	✅ DO: Clearly structure content (headings, summaries, lists). ✅ DO: Offer compact, copyable answers. ❌ DON'T: Use redundant or repetitive content. ❌ DON'T: Add unnecessary filler or fluff.

‍

Key Takeaways

‍

Freshness Matters: Regularly updated and clearly dated content is essential.
Be Unique and Niche: Region-specific, personalised, or highly specialised content improves chances of citation.
Interactive Content: Calculators, planners, and dynamic databases significantly boost citation likelihood.
Clarity and Structure: Keep your content structured, quotable, and easy to skim.
SEO Implications: Traditional keyword optimisation is less relevant; the focus should shift to creating factual, clear, and modular content suitable for AI citation.
Clicks and Visitors: If your business model relies on traffic, ensure your content is not only quotable but also link-worthy.

‍

Full credit for these insights goes to the original, in-depth article from GPT-Insights—definitely worth a read for anyone serious about optimising content visibility for AI models.

‍

AI Engine	Favoured Sources & Patterns	Works Best For …	Optimisation Tips
ChatGPT (OpenAI)	Authoritative “reference” web: Wikipedia ≈ 27 % of citations; high-authority news brands, peer-reviewed research. UGC almost absent (< 0.5 %).	B2B & technical explainer content where verified facts matter.	• Publish fact-rich hub pages with clear sourcing. • Win mentions in major news outlets & specialist journals (these surface often).(netinfluencer.com)
Claude (Anthropic)	Defaults to internal knowledge; will cite only when asked. When it does, leans to official docs, standards bodies, gov datasets. Very low UGC share (similar to ChatGPT).	Regulated niches (finance, health, compliance) that demand primary data.	• Add inline citations / footnotes in your own pages—Claude can surface them verbatim. • Provide downloadable PDFs/white-papers (Claude cites uploaded docs readily).(docs.anthropic.com)
Google AI Overviews / Gemini	Strongest appetite for UGC: Reddit is #1 single domain; Quora & LinkedIn posts appear. UGC ≈ 4 % (Overviews) / 2 % (Gemini) of all citations. 82 % of links point to deep sub-pages.	B2C discovery queries (“best X”, “top Y”).	• Seed authentic threads on Reddit / Quora. • Optimise deep pages with schema & jump-links (Google rarely cites homepages).(netinfluencer.com, searchengineland.com)
Perplexity AI	Middle ground: UGC ≈ 1 %; 60 % of citations overlap Google top-10. Pulls FAQs, official docs, reputable blogs.	Mixed-intent queries, research comparisons.	• Rank organically: top-10 Google visibility correlates with Perplexity links. • Create concise answer blocks (Perplexity loves copy-pasting summaries).(netinfluencer.com, searchengineland.com)
Microsoft Copilot / Bing Chat	(Not in Rankscale study) Pattern mirrors Bing results: heavy on Microsoft-indexed news & docs; cautious on UGC.	Tech, dev, software “how-to” queries.	• Mark up docs with FAQ schema. • Provide code snippets—Copilot pulls these directly.

‍

How to use the matrix

‍

B2C brands: focus on community & review signals for Google’s engines; still maintain authority pages for ChatGPT/Claude.
B2B / technical: invest in white-papers, standards citations and analyst coverage—ChatGPT, Claude, Perplexity cite those most.
Universal play: build deep, well-structured sub-pages; win third-party coverage (news, Wikipedia); nurture authentic forum threads.

‍

Updated as of 3rd July 2025 - Google AI Mode's Volatility & Source Patterns.

‍

Only 9% of Google AI Mode URLs Repeat

‍

SE Ranking recently published a deep-dive study into Google’s AI Mode, revealing just how volatile and source-agnostic it can be compared with traditional search and AI Overviews. Here are the key takeaways every Aussie marketer should know (even tho the test is running in the US market):

‍

Extreme Volatility: Only 9.2 % of exact URLs overlapped across three separate AI Mode searches for the same query. In other words, AI Mode “re-fans out” your query each time, often returning completely different link sets—even on the same day.
Trusted Domains Dominate: Despite low URL overlap, most citations still come from familiar, high-authority sites (e.g., Indeed, Wikipedia, Reddit, YouTube)—indicating AI Mode leans on trusted sources rather than chasing the latest ranking algorithm.
Independent of Organic Rank: With just 9.2 % URL overlap, AI Mode’s selections appear largely independent of Google’s organic top-10 results, suggesting a separate “search quota” and citation process distinct from classic ranking.

‍

Search Engine Journal also weighed in, noting that these findings underscore a shift in how AI search features deliver answers compared to traditional AI Overviews, where overlap with organic results remains low but slightly higher than in AI Mode

‍

What This Means for Your Strategy

‍

Diversify Your Footprint: Don’t rely solely on top-ranked pages—build authority across multiple, high-trust domains.
Monitor AI Volatility: Track your core queries in AI Mode regularly; expect different “winners” each session.
Optimise for Trust: Emphasise clear sourcing and domain authority in your content, since AI Mode citations favour established sites.

‍

While this research focuses on Google’s AI Mode, it hints at similar patterns emerging in other AI-powered search tools. Prepare by broadening your content distribution, reinforcing domain credibility, and staying agile as AI search evolves.

‍

‍

References

https://gpt-insights.de/ai-insights/gpt-insights-claude-leak-en/
https://seranking.com/blog/ai-mode-research/
https://www.searchenginejournal.com/study-google-ai-mode-returns-largely-different-results-across-sessions/550249/

‍

Add DEANLONG.io to your Google News feed

Send Dean A Message If You Have Any Question!

Best Practices

Industry Update

SEO

Trends

Hongxin(Dean) Long

Dean Long is a Sydney-based performance marketing and communication professional with expertise in paid search, paid social, affiliate, and digital advertising. He holds a Bachelor's degree in Information Systems and Management and is also a distinguished MBA graduate from Western Sydney University.

Your Cart