How to Optimise Your Blog for Google AI Overviews and ChatGPT Citations
Search is changing from a list of blue links into an answer-driven experience. Google AI Overviews and AI Mode, ChatGPT Search, Perplexity and other AI-powered discovery systems can synthesise information from multiple sources and send users directly to pages that support the answer. For publishers, the opportunity is no longer simply to rank for a keyword. The bigger opportunity is to become a reliable source that an answer engine can discover, understand, trust and cite.
Table of Contents
- 1. Why AI Search Changes SEO
- 2. Understanding AI Overviews, AI Mode and AI Search
- 3. What Makes a Website Citation-Worthy?
- 4. Build Content That Is Easy to Understand
- 5. The Direct-Answer Content Framework
- 6. Original Research and First-Hand Experience
- 7. E-E-A-T, Authors and Trust
- 8. Technical SEO for AI Search
- 9. Structured Data: What Actually Matters
- 10. Internal Linking and Topical Authority
- 11. Images, Video and Visual Evidence
- 12. Preparing Your Site for ChatGPT Search
- 13. Preparing Your Site for Perplexity
- 14. Measuring AI Search Visibility
- 15. Common GEO and AEO Mistakes
- 16. A Practical 30-Day Implementation Plan
- 17. Frequently Asked Questions
- 18. Conclusion
1. Why AI Search Changes SEO
Traditional search optimisation was built around a relatively simple objective: understand what people search for, create a relevant page, make the page accessible to search engines, build authority and earn a position in the search results.
AI-powered search adds another layer. A user can now ask a long, conversational question instead of entering a short keyword. The system may break that question into several related searches, retrieve information from different websites, compare the evidence and generate a consolidated answer.
This does not mean that traditional SEO has become obsolete. Quite the opposite. Google states that its AI search experiences are rooted in its existing Search systems and that the fundamentals of SEO continue to matter. Pages still need to be crawlable, indexable, useful, relevant and technically accessible.
What changes is the competitive environment. Your article may no longer compete only for a position in a conventional result page. Individual passages, facts, explanations, tables, examples and original research may become evidence used to construct an AI-generated response.
The strategic shift: Stop asking only, “How do I rank this article?” Start asking, “If an AI system needs a trustworthy answer to this question, what would make my page one of the best sources available?”
That question leads to a more sustainable form of optimisation: excellent information architecture, strong technical SEO, first-hand expertise, original evidence, transparent authorship and content that directly satisfies the user's underlying need.
2. Understanding AI Overviews, AI Mode and AI Search
AI search is not one single technology. Different products use different retrieval systems, ranking signals, models and presentation formats. Therefore, publishers should avoid assuming that one universal “GEO algorithm” exists.
Google AI Overviews
Google AI Overviews can provide an AI-generated summary for queries where Google determines that an additional synthesis is useful. Supporting links allow users to investigate the underlying information.
The important practical point is that eligibility still depends on the page being accessible and eligible for Google Search. Google does not require publishers to create a special “AI Overview schema” or an AI-specific content format.
Google AI Mode
AI Mode is designed for more exploratory and complex searches. A user may ask for comparisons, explanations or multi-step research rather than a single fact.
This makes comprehensive topic coverage particularly valuable. A strong article should answer the primary question while also providing enough context to help the reader make a decision or understand the subject.
ChatGPT Search
ChatGPT can search the web and provide answers with links to relevant sources. For publishers, this creates an additional discovery channel where being crawlable and clearly attributable can matter.
OpenAI currently identifies OAI-SearchBot as the crawler associated with helping websites appear in ChatGPT search results. Publishers who want their pages discoverable through ChatGPT should therefore review their robots.txt and hosting configuration rather than relying on speculative “AI SEO hacks.”
Perplexity
Perplexity is strongly oriented around web research and source attribution. Its current documentation states that PerplexityBot respects robots.txt directives and that its search experience can display sources supporting generated answers.
| Platform | What Publishers Should Prioritise | Practical Focus |
|---|---|---|
| Google AI Overviews | Strong Google Search fundamentals | Helpful content, crawlability, relevance, authority and clear page structure |
| Google AI Mode | Comprehensive topic coverage | Useful explanations, comparisons, evidence and supporting content |
| ChatGPT Search | Discoverability and source quality | Crawlability, clear attribution and useful original information |
| Perplexity | Research-grade usefulness | Accurate claims, transparent authorship and supporting evidence |
3. What Makes a Website Citation-Worthy?
There is no guaranteed formula for obtaining an AI citation. A page can be technically perfect and still not be selected. Conversely, a relatively small website can become highly visible when it contains information that larger publishers do not provide.
The most useful way to think about citation-worthiness is through five characteristics:
- Relevance: The page directly addresses the user's question.
- Accuracy: Claims are supported by reliable evidence or clearly identified first-hand experience.
- Specificity: The article contains useful details rather than generic statements.
- Originality: The publisher contributes something that cannot simply be copied from ten competing articles.
- Attribution: Readers can identify who created the information and why that person or organisation is qualified to discuss it.
The Citation-Worthy Content Test
Before publishing an important article, ask: “What information exists on this page that would make another researcher comfortable citing it?”
If the honest answer is “nothing beyond what everyone else has already published”, the article probably needs deeper research, first-hand testing, original examples or a stronger point of view.
4. Build Content That Is Easy to Understand
One of the biggest mistakes in modern SEO is confusing “machine-readable” with “machine-written”. You do not need artificial wording, excessive headings or robotic paragraphs to make content understandable.
The best approach is to make the page exceptionally clear for humans. Good human information architecture generally produces content that search systems can also interpret efficiently.
Use a logical hierarchy
Your article should have one clear primary topic. H2 headings should represent major concepts, while H3 headings should explain meaningful subtopics.
Avoid headings that exist only to insert keywords. A heading such as “Best AI SEO Strategies for AI SEO Search Optimisation” looks artificial. A heading such as “How to Make Technical Tutorials Easier to Discover” communicates a genuine information need.
Write short, information-rich paragraphs
Short paragraphs improve scanning, especially on mobile devices. However, do not turn every sentence into a separate paragraph simply because you have heard that “AI prefers chunks”. There is no Google requirement for a particular paragraph length.
Instead, group sentences around a single idea. A paragraph should normally answer one small question, explain one relationship or provide one piece of evidence.
Use lists when the information is naturally list-shaped
Lists work well for steps, requirements, comparisons, advantages, limitations and checklists. They should not be inserted mechanically into every article.
5. The Direct-Answer Content Framework
A useful editorial technique is to answer the reader's immediate question early, then provide supporting detail. This is sometimes described as an inverted-pyramid structure.
However, avoid rigid claims such as “AI systems always extract 40–60 words under every H2”. There is no reliable Google rule that says this. The better principle is simple: put the important answer where a reader can find it quickly.
A practical structure
- Answer: Give the direct conclusion.
- Context: Explain why it is true.
- Evidence: Provide data, examples, documentation or testing.
- Limitations: Explain when the conclusion may not apply.
- Action: Tell the reader what to do next.
Example
Instead of beginning a technical explanation with several paragraphs of background, start with the practical answer:
Example: “A slow WordPress page is not automatically caused by your hosting provider. Large images, third-party scripts, excessive plugins, render-blocking resources and poor caching can all contribute to slow page delivery. Diagnose the bottleneck before replacing the host.”
The following sections can then explain how to diagnose each cause.
6. Original Research and First-Hand Experience
This is one of the strongest opportunities for smaller publishers.
A small technology blog cannot easily outspend a multinational publisher on content production. It can, however, publish an experiment that nobody else has performed.
Examples of original evidence
- Benchmarking two web frameworks under identical conditions.
- Testing website performance on several budget Android devices.
- Comparing page speed before and after a specific optimisation.
- Documenting the results of a real migration from one hosting provider to another.
- Testing how different image formats affect loading performance.
- Publishing anonymised results from a real SEO experiment.
- Creating an original dataset and explaining its methodology.
- Building a reproducible coding example and publishing the source code.
The key word is reproducible. If you publish a benchmark, explain the environment. If you report a performance improvement, explain what changed. If you make a technical recommendation, explain how you reached it.
Example benchmark methodology
Test environment
---
Device: Android smartphone
Browser: Chrome
Network: 4G
Page type: Article
Image format: WebP
Cache: Enabled
Test procedure
1. Clear browser cache.
2. Load the page three times.
3. Record the median result.
4. Repeat after optimisation.
5. Compare the median values.
A methodology like this transforms a generic claim into an evidence-backed observation.
7. E-E-A-T, Authors and Trust
Experience, expertise, authoritativeness and trustworthiness are important concepts when evaluating content quality. They are especially relevant to technology publishing because technical advice can affect security, privacy, finances, software deployments and business decisions.
Show who created the content
A professional technology publication should make authorship understandable. Depending on the organisation, this can include:
- Author name.
- Short professional biography.
- Relevant qualifications or practical experience.
- Links to appropriate professional profiles.
- Editorial review information where applicable.
- Publication and update dates.
- A clear correction policy.
Do not manufacture expertise
Never create fictional experts, invented quotations or fake testing results simply to make an article appear authoritative.
The original article's fictional quotation from a “Senior Search Architect & AI Information Retrieval Specialist” is a good example of what should be removed unless an actual, identifiable expert supplied that statement.
Authentic authority comes from evidence, experience and transparency—not from impressive-sounding job titles.
Keep claims proportional to evidence
Avoid statements such as “This technique guarantees AI citations” or “Google's AI always prefers 50-word answers”. Search systems are probabilistic and constantly evolving.
Strong technical writing distinguishes between:
- Documented fact — supported by official documentation.
- Observed result — supported by your own experiment.
- Expert interpretation — a reasoned conclusion.
- Hypothesis — an idea that still requires validation.
8. Technical SEO for AI Search
Technical SEO remains the foundation. Google states that pages need to meet the normal technical requirements for Search to be eligible as supporting links in AI features.
1. Make important pages crawlable
Review robots.txt, meta robots directives, canonical URLs, redirects and server responses. A page cannot contribute effectively to search visibility if the relevant crawler cannot access it.
2. Avoid accidental noindex directives
Check important articles for accidental noindex tags. This is particularly important after theme changes, migrations and plugin installations.
3. Make the primary content available as text
Important information should not exist exclusively inside an image, canvas element or inaccessible interactive component.
Images are valuable, but the essential explanation should also be represented in HTML text.
4. Use clean canonicalisation
If several URLs represent substantially the same content, use appropriate canonicalisation and avoid unnecessary duplication.
5. Improve page experience
Mobile usability, loading performance, layout stability, readable typography and unobtrusive advertising all contribute to a better publishing environment.
6. Do not overcomplicate JavaScript
Modern search engines can process JavaScript, but excessive client-side rendering can create unnecessary complexity. Whenever practical, deliver important content in accessible HTML.
Technical AI-Search Checklist
- Important pages return a successful HTTP response.
- Robots.txt does not unintentionally block important crawlers.
- Important pages are indexable.
- Canonical URLs are logical and consistent.
- Main content is available in text.
- Internal links connect related articles.
- Pages work well on mobile devices.
- Images are compressed and appropriately described.
- Structured data accurately reflects visible content.
9. Structured Data: What Actually Matters
Structured data can help search engines understand the entities and content types represented on a page. But structured data should not be treated as a secret mechanism for forcing AI citations.
Google explicitly states that there is no special structured data requirement for AI Overviews or AI Mode. Existing structured-data guidelines still apply where a particular Search feature supports that markup.
Article schema
For a blog article, appropriate Article structured data can communicate information such as headline, author, image and publication details.
Breadcrumb schema
Breadcrumb markup can help communicate the site's hierarchy and page position.
FAQ structured data
FAQ content can be useful for readers, but publishers should not add FAQ markup simply because they believe it guarantees an AI feature. Structured data must represent visible, accurate page content and comply with the relevant Google guidelines.
HowTo structured data
For instructional content, clearly structured steps are valuable. However, publishers should verify the current Google Search support and eligibility requirements for any structured-data feature before implementing it.
Important: Never add structured data for information that is not genuinely present on the page. Structured data is not a place to insert hidden keywords, fake ratings, fictional authors or unsupported claims.
The safest implementation philosophy is simple: mark up what is genuinely on the page, keep it accurate and validate it before deployment.
10. Internal Linking and Topical Authority
A high-quality article should not exist as an isolated URL. Build a connected knowledge structure.
Imagine a technology blog publishing an article about AI SEO. It could connect that article to supporting pages covering technical SEO, schema markup, Core Web Vitals, keyword research, content strategy, internal linking and Google Search Console.
These relationships help readers discover related information and help search engines understand the broader subject matter of the website.
Use contextual internal links
Instead of writing “Click here”, use descriptive anchor text that explains the destination.
For example:
Weak: Learn more
Better: Learn how to audit Core Web Vitals
Create hub-and-spoke content
A practical structure is:
- Hub: Complete Guide to Technical SEO
- Spoke: Core Web Vitals Guide
- Spoke: Schema Markup Guide
- Spoke: Internal Linking Guide
- Spoke: JavaScript SEO Guide
- Spoke: Image SEO Guide
Each supporting article should provide genuine value rather than existing only to create links.
11. Images, Video and Visual Evidence
AI search is not limited to plain text. Google has specifically highlighted opportunities involving images and video within its guidance for generative search.
For technology publishers, visual content can strengthen an article when it explains something that is difficult to communicate through text alone.
Useful technology visuals include:
- Architecture diagrams.
- Before-and-after performance comparisons.
- Annotated screenshots.
- Benchmark charts.
- Workflow diagrams.
- Code-output demonstrations.
- Product comparison tables.
- Original infographics.
Avoid decorative images that merely occupy space. A useful image should clarify, demonstrate or strengthen the information on the page.
Give images descriptive filenames and appropriate alternative text. The alt text should describe the image's meaningful content rather than functioning as a keyword list.
12. Preparing Your Site for ChatGPT Search
ChatGPT search introduces a practical publisher consideration: crawler access.
OpenAI's current publisher guidance says websites should avoid blocking OAI-SearchBot if they want their content to be discoverable, surfaced and clearly cited in ChatGPT search results.
This does not mean allowing every automated crawler. Website owners should distinguish between different crawler purposes and make deliberate choices based on their publishing and licensing policies.
Review robots.txt
Your robots.txt configuration should be intentional. A simple technical audit should determine:
- Which crawlers are allowed?
- Which directories are blocked?
- Are important article URLs accessible?
- Is a CDN or firewall blocking legitimate crawlers?
- Are rate limits producing repeated 403 or 429 responses?
Track referral traffic
ChatGPT referrals can be analysed in analytics platforms. OpenAI's current publisher documentation notes that ChatGPT search referral URLs can include a utm_source=chatgpt.com parameter, which can help publishers identify this traffic.
Treat this traffic like any other acquisition channel. Measure landing pages, engagement, conversions and returning users rather than focusing only on raw visits.
13. Preparing Your Site for Perplexity
Perplexity's search experience is particularly relevant to publishers that produce explanatory and research-oriented content because its answers commonly expose supporting sources.
Perplexity's current documentation states that PerplexityBot respects robots.txt. If your content is intentionally blocked, you should not expect the crawler to index the page's full or partial text.
Make the evidence easy to verify
If you publish a claim, provide the context needed to evaluate it.
For example, rather than writing:
“This hosting provider is 40% faster.”
write something closer to:
“In our test of the same WordPress installation under the conditions described below, the median page-load measurement improved by approximately 40% after enabling server-side caching. This result should not be interpreted as a universal performance advantage because hosting performance varies by configuration.”
The second statement is more useful because it communicates methodology, scope and limitations.
14. Measuring AI Search Visibility
Measuring generative visibility is more complicated than checking one ranking position.
Google Search Console
Search Console remains an important source of search performance information. Google has also introduced reporting and documentation around performance in its generative search experiences.
Monitor:
- Impressions.
- Clicks.
- Click-through rate.
- Queries.
- Top landing pages.
- Changes after major content updates.
Analytics
Track referral traffic from AI interfaces where identifiable. More importantly, measure what happens after the visitor arrives.
- Engaged sessions.
- Average engagement time.
- Newsletter subscriptions.
- Affiliate clicks.
- Tool usage.
- Contact enquiries.
- Purchases or other conversions.
Manual citation monitoring
Periodically test important questions in relevant AI search products. Use the same query categories that your audience actually asks.
Record:
- The question.
- The date tested.
- Whether your site appeared.
- Which page was cited.
- Which claim was referenced.
- Which competitors appeared.
- What evidence competitors provided that your page lacked.
Do not treat a single AI response as a permanent ranking. AI-generated search results can change based on query wording, location, freshness, source availability and other factors.
15. Common GEO and AEO Mistakes
Mistake 1: Creating “AI-only” content
Content should primarily serve the reader. Google explicitly recommends people-first, useful and reliable content rather than pages created mainly to manipulate search or generative systems.
Mistake 2: Creating dozens of nearly identical pages
Publishing separate articles for every tiny variation of a query can create thin or repetitive content. Cover meaningful variations naturally within a strong resource where appropriate.
Mistake 3: Keyword stuffing
Repeating phrases such as “AI SEO”, “GEO”, “ChatGPT citations” and “Google AI Overview” in every paragraph makes content harder to read and does not transform it into authoritative material.
Mistake 4: Inventing statistics
Never create percentages, benchmark figures, survey results or performance claims simply because they make an article look more authoritative.
Mistake 5: Fake experts
Do not fabricate expert quotations or attribute generic statements to imaginary professionals.
Mistake 6: Treating schema as a ranking switch
Structured data helps search engines understand eligible content types and can support certain search features, but valid markup does not guarantee a rich result or AI citation.
Mistake 7: Obsessing over “AI word counts”
There is no magic article length that guarantees AI visibility. A 700-word page can outperform a 4,000-word page if it provides a better answer.
Mistake 8: Ignoring update discipline
Technology changes quickly. An article about an API, browser feature, AI model, security practice or software release can become misleading if old information remains untouched.
Editorial rule: Never add “Updated 2026” to an article unless you actually reviewed and improved the content. A changed date without substantive review weakens reader trust.
16. A Practical 30-Day Implementation Plan
You do not need to rebuild an entire website in one weekend. A systematic month-long programme is more realistic.
Week 1 — Technical audit
- Verify Google Search Console.
- Review indexing problems.
- Inspect robots.txt.
- Check important canonical URLs.
- Test mobile usability.
- Identify slow templates.
- Remove accidental noindex directives.
- Review broken internal links.
Week 2 — Upgrade your best articles
Select five to ten articles that already receive impressions or traffic.
For each article:
- Rewrite the opening to answer the searcher's primary question.
- Remove outdated claims.
- Add original examples.
- Add supporting evidence.
- Improve headings.
- Add useful internal links.
- Improve images.
- Review author and update information.
Week 3 — Build original authority
Publish at least one piece of genuinely original research.
It does not need to be a massive industry study. A carefully designed technical experiment can be extremely useful.
For example:
Original Research Idea
“We tested five image formats on a Blogger technology article using identical dimensions, compression targets and network conditions. Here are the file sizes, visual differences and loading observations.”
That article has a much stronger reason to exist than another generic “10 image SEO tips” post.
Week 4 — Measure and refine
- Review Search Console performance.
- Review analytics referral data.
- Check important queries in AI search products.
- Identify articles competitors cite more frequently.
- Improve weak sections.
- Build additional supporting articles where a genuine content gap exists.
17. The Premium AI-Search Content Template
For future Tech Reflector articles, the following editorial framework can be reused without turning every article into an identical template.
Recommended Structure
- Clear title: Describe exactly what the reader will learn.
- Strong introduction: Explain the problem and why it matters.
- Direct answer: Address the primary intent early.
- Context: Explain the underlying technology.
- Practical implementation: Show the reader what to do.
- Evidence: Include documentation, testing or original observations.
- Limitations: Explain where the recommendation does not apply.
- Examples: Use realistic technical scenarios.
- Visuals: Add diagrams, screenshots or charts where they genuinely help.
- Internal links: Connect related knowledge.
- FAQs: Answer legitimate remaining questions.
- Conclusion: Give the reader a practical next step.
18. Frequently Asked Questions
No. Google states that its normal SEO fundamentals remain relevant for AI features and that there are no additional technical requirements or special AI-specific schema types required for inclusion in AI Overviews or AI Mode. Focus on crawlability, helpful content, strong technical SEO, clear structure and genuine value.
No. For Google Search, generative experiences are built on top of core Search systems. Strong technical SEO, relevant content, crawlability, internal linking, authority and user experience remain foundational.
No. Concise answers can improve usability, but there is no universal word-count rule that guarantees an AI citation. Write enough to answer the question accurately and provide the evidence and context required.
Do not treat llms.txt or similar AI-specific files as a required Google ranking mechanism. Google's current guidance says publishers do not need to create new machine-readable files to appear in its generative AI search features.
The important issue is not simply whether an AI tool assisted with drafting. Content should meet applicable Search policies and provide genuine value. Automatically generated material that is scaled, repetitive, misleading or created primarily to manipulate rankings can create serious quality and spam risks.
Yes, particularly where the smaller publisher has first-hand experience, original testing, niche expertise or information that larger sites do not provide. A smaller website should compete on usefulness and specificity rather than trying to imitate a large media organisation.
Make your content publicly accessible and technically crawlable, avoid unintentionally blocking relevant search crawlers, publish accurate and original information, identify authors clearly, and provide useful pages that genuinely answer questions. OpenAI currently advises publishers not to block OAI-SearchBot when they want content to be discoverable in ChatGPT search.
No. Structured data helps search engines understand eligible content and can support certain search features, but valid structured data does not guarantee a rich result, higher ranking or AI citation.
Update frequency should depend on how quickly the topic changes. Software, AI models, APIs, security practices and product specifications may require frequent reviews, while evergreen concepts can remain accurate much longer. Update an article when its facts, examples, links or recommendations genuinely need improvement.
19. Conclusion: Build Sources, Not Just Pages
The biggest mistake publishers can make in the AI-search era is treating generative search as another ranking trick.
It is better understood as an evolution in information discovery.
Search engines and AI assistants increasingly have the ability to synthesise information from multiple sources. That creates a new competitive advantage for publishers that produce information worth synthesising.
The strongest strategy is therefore surprisingly familiar: understand your audience, solve real problems, publish accurate information, demonstrate experience, make the site technically accessible, organise related knowledge intelligently and continually improve your work.
Then add the capabilities that matter in an AI-mediated search environment: clear answers, strong evidence, original research, transparent authorship, useful multimedia, logical information architecture and measurable outcomes.
The 10 Rules to Remember
- Do not abandon traditional SEO for GEO.
- Create content for people first.
- Give the answer early when the user needs a direct answer.
- Support important claims with evidence.
- Publish original research whenever possible.
- Demonstrate real experience rather than manufacturing authority.
- Keep technical SEO and crawlability strong.
- Use structured data accurately, not as a ranking trick.
- Build topical depth through genuinely useful internal linking.
- Measure actual visibility, traffic and conversions instead of chasing hypothetical AI rankings.
Ultimately, the objective is not to “hack” Google AI Overviews or force ChatGPT to mention your website. The objective is much more durable: become the source that deserves to be mentioned.
In the era of generative search, the winners will not necessarily be the websites that publish the most content. They will be the publishers that consistently provide the clearest answers, strongest evidence and most useful original knowledge.
