In short: this checklist gathers 25 actions for reviewing a brand’s presence in answers generated by ChatGPT, Claude, Gemini and Perplexity. It covers technical foundations, content, trust signals, platforms and measurement. It does not guarantee rankings or citations: use it to prioritize checks, document changes and compare observations over time. See the complete AEO audit guide.
Search products and conversational assistants do not expose the same sources or answer formats. This checklist helps review what is accessible, understandable and measurable on a site, without assigning an observed result to a single cause.
Starting point
Before changing anything, record the pages concerned, the queries or questions being tracked, the observation date and the method used. Keep a measured change; verify an isolated variation before drawing conclusions.
This checklist is not a wish list. It's a structured action plan in 5 categories, with 5 actions per category. Each action comes with a clear explanation and an immediate implementation tip.
How to use this checklist
Work through the five categories in the order that matches your observed gaps. Technical foundations make pages accessible and verifiable; content, evidence and measurement then document what changes. No section guarantees a citation.
Check off each action once implemented. Revisit this list monthly to maintain and improve your AI visibility.
Our starting point: separate what a tool can measure from what a change can actually establish. The Discovery plan limits and invented-URL observations below do not prove that any checklist action works.
Verified reference points from AI Labs Audit
| Reference point | Scope checked on November 9, 2026 |
|---|---|
| Discovery plan | 3 AI systems × 5 scenarios per audit; 100 monthly credits and 500 welcome credits. |
| Consultant plan | 1,500 monthly credits. The pricing table lists a catalog of 50+ available models. |
| Observatory: invented URLs | 70 invented URLs out of 26,435 checked URLs, or 0.26%. |
Our view: checking cited URLs is more defensible than promising a citation from markup. The Observatory aggregates audits of brands that chose to be measured, not a representative sample of all businesses. Keep the scope constant when comparing observations.
Pricing and plan limits ; Observatory and sample limitations
1. Technical fundamentals
Check page accessibility and descriptions
#1 Implement Schema.org markup (Article, Organization)
Why it matters
Structured data describes a page type and its entities. It must match visible content. Validate the vocabulary with the Schema Markup Validator; use Google’s Rich Results Test only for rich result types it still supports.
Implementation tip
Start with Organization on the homepage and Article on articles. HowTo remains a Schema.org type, but Google has retired its rich results. Valid markup does not guarantee display.
#2 Assess the value of an llms.txt file
Why it matters
The llms.txt file is an emerging convention. Its measurable effect on AI visibility is not established; assess it as an optional complement for your content and users.
Implementation tip
If you choose to implement it: a simple text file at the root (yoursite.com/llms.txt) is enough. Keep it up to date.
#3 Optimize page speed and stability
Why it matters
A page that renders reliably and remains readable removes access friction; speed alone does not establish presence in an AI answer.
Implementation tip
Use PageSpeed Insights and Lighthouse on important templates, then fix heavy resources and rendering errors.
#4 Ensure a mobile-first design
Why it matters
Google uses mobile-first indexing, and readable mobile interfaces help visitors consult content.
Implementation tip
Check text, viewport, touch interactions and horizontal scrolling.
#5 Configure robots.txt correctly
Why it matters
Training bots and search bots do not have the same role. robots.txt rules should reflect an explicit decision for each bot family.
Implementation tip
Document GPTBot separately from OAI-SearchBot, and ClaudeBot separately from Claude-SearchBot; review official documentation and your objectives before setting rules.
Keep crawling separate from indexing and citation: a bot visit does not prove either of the latter.
2. Content structure
Organize answers to make reading easier
#6 Use the "answer-first" format
Why it matters
A direct answer helps readers understand the topic quickly. Any effect in an assistant depends on the query, system and available sources.
Implementation tip
Answer the main question in the introduction, then add definitions, limits and examples.
Answer engines may use different sources; a clear answer is useful to readers first.
#7 Add a clear H2/H3 hierarchy
Why it matters
A logical H1, H2 and H3 hierarchy helps readers navigate and assistive technologies move through a page. It does not establish a universal mechanism for AI extraction or citation.
Implementation tip
Each H2 should answer a sub-question. Formulate them as questions or clear statements. Example: "How does AEO work?" rather than "AEO".
#8 Include FAQ sections
Why it matters
A visible FAQ can address real reader questions. It does not guarantee display or reuse by a search product or assistant.
Implementation tip
Add only useful questions. FAQPage can describe a visible FAQ, but Google has not shown FAQ rich results since May 7, 2026. This markup does not guarantee an AI citation.
#9 Create numbered and bulleted lists
Why it matters
Lists make a process or comparison easier to scan. They do not guarantee reuse or display in a search product or assistant.
Implementation tip
Transform your dense paragraphs into lists when possible. Use numbered lists for processes/steps, bullets for features/benefits.
#10 Add TL;DR summaries
Why it matters
An introductory summary helps readers understand the essential point without reading the whole page. It is not a citation guarantee.
Implementation tip
Place a short block with the main answer, its scope and one useful limitation.
3. Authority and E-E-A-T
Document authors, sources and evidence
#11 Create detailed author pages
Why it matters
An author page helps identify where content comes from and how a reader can assess its context.
Implementation tip
Name the author, verifiable experience, contributions and context without overstating credentials.
#12 Add proof of credentials and experience
Why it matters
Verifiable examples, dated data and explicit limits make content easier for readers to check.
Implementation tip
Add case studies only with consent and evidence; distinguish observation, hypothesis and recommendation.
#13 Build thematic authority clusters
Why it matters
Interlinked pages on one topic aid navigation and reduce repetition; they do not alone prove authority.
Implementation tip
Link a reference page to complementary articles and check that every link reaches a useful page.
#14 Get external citations and mentions
Why it matters
Third-party mentions let readers find independent sources. They do not guarantee visibility in generated answers.
Implementation tip
Prefer useful editorial contributions and accurate profiles; avoid link exchanges and artificial reviews.
#15 Keep content current when facts change
Why it matters
A date alone does not make a page reliable. Update it when a fact, source, product or recommendation has materially changed.
Implementation tip
Review pages after a verifiable change and remove uncertain figures or links.
4. Multi-platform optimization
Check information across relevant channels
#16 Test your visibility on ChatGPT, Perplexity, and Gemini
Why it matters
Assistants can return different answers to the same question. Observe them separately rather than generalizing one result.
Implementation tip
Keep the same question set and record date, mode and displayed sources where available.
When a system uses RAG, it retrieves documents before generating an answer; distinguish this from a response without document retrieval.
#17 Monitor Google AI Overviews
Why it matters
AI Overviews do not appear for every query and can vary by country, device and observation time.
Implementation tip
Search priority queries in a defined context and compare only comparable observations.
#18 Optimize for voice search
Why it matters
Voice queries are often complete questions. Clear answers can help readers without guaranteeing voice distribution.
Implementation tip
Write simple answers to natural questions; use Speakable only within Google’s documented scope.
#19 Build a presence on Reddit and Quora
Why it matters
Community discussions can add useful context, but their existence does not guarantee AI reuse.
Implementation tip
Participate when you can provide a verifiable answer and follow community rules.
#20 Register in relevant industry directories
Why it matters
An accurate directory profile can help prospects verify a business. It does not prove more visibility for comparison queries.
Implementation tip
Choose directories genuinely used in your sector and keep verifiable information consistent.
Check NAP consistency (name, address and phone) where these details are relevant.
5. Measurement and continuous improvement
Track your progress and optimize continuously
#21 Set up AI visibility tracking
Why it matters
AI visibility tracking documents observed answers, sources and mentions within a defined scope. On its own, it cannot attribute a change to one specific action.
Implementation tip
Use specialized tools like AI Labs Audit. Track referral traffic from AI (chatgpt.com, perplexity.ai) in Google Analytics. Set up alerts on your mentions.
#22 Define your strategic questions to monitor
Why it matters
Choose a limited, representative set of business questions phrased as prospects ask them, then keep that set stable for comparisons.
Implementation tip
List: your customers' pre-purchase questions, comparisons with competitors, "best X for Y" queries, questions about your sector. Categorize by business priority.
#23 Calculate your AI Share of Voice
Why it matters
AI Share of Voice is the number of mentions of your brand divided by the total mentions of all brands being compared, multiplied by 100 to obtain a percentage. Brand mentions are distinct from citations of source URLs. This is neither an overall market share nor a search-ranking equivalent.
Implementation tip
Record each brand’s mentions separately from URLs cited as sources. To compare observations, keep the same brands, questions, models, modes and protocol, and use comparable observation periods.
#24 Compare content formats with a documented method
Why it matters
A before-and-after comparison may reveal change, but it is not an A/B test and cannot alone demonstrate causality.
Implementation tip
Keep the baseline, change one identified element, record dates and other changes, then treat gaps as observations.
#25 Produce monthly AI visibility reports
Why it matters
Regular reporting makes decisions and observations traceable; it does not turn a variation into proof.
Implementation tip
Include scope, questions, dates, observed answers or citations, published changes and limits.
Summary: the 5 categories at a glance
- Technical (1-5): Schema.org, llms.txt, speed, mobile, robots.txt
- Content (6-10): Answer-first, H2/H3, FAQ, lists, TL;DR
- Authority (11-15): Author pages, credentials, clusters, citations, freshness
- Multi-platform (16-20): AI tests, AI Overviews, voice, Reddit/Quora, directories
- Measurement (21-25): Tracking, key questions, SOV, documented comparisons, reports
Frequently asked questions
What is an AEO checklist?
An AEO checklist is a set of checks and actions for reviewing a brand’s presence in conversational-assistant answers. It organizes work across technical foundations, content, trust signals and observation without promising a particular outcome.
Which AEO actions should come first?
Start by documenting the baseline, checking access and structured data on important pages, then clarifying answers to prospective customers’ questions. Priority depends on the observed gap, not on a universal sequence.
How can I check whether a site is ready for AEO?
Keep a representative set of questions, observe answers in the assistants your audience uses, and record the context for each result. Also check mobile rendering, indexability and structured data with appropriate tools.
Do I need a large budget to begin?
No. A first cycle can rely on an inventory of pages, improvements to existing answers and regular observation. The budget then depends on content volume, number of markets and the depth of monitoring required.
How often should I review an AEO checklist?
Review it after meaningful changes to pages, structured data or the questions being tracked. A regular check, for example monthly, helps compare observations when the method and scope stay consistent.
Downloadable checklist with proof, effort and owner
Use this table as the working version of the checklist: each action names the proof you keep, the effort it takes and who is responsible. Print it or copy it into your project tool.
| Action | Proof | Effort | Owner |
|---|---|---|---|
| robots.txt allows AI bots (GPTBot, ClaudeBot, PerplexityBot) | Screenshot of robots.txt + crawl test | Low | Technical |
| Structured data (Schema.org) on key pages | Validator report | Medium | Technical |
| FAQ marked up with the AQA standard | Rendered FAQPage markup | Medium | Content |
| Clear question-shaped headings on priority pages | Page extract before and after | Low | Content |
| Author and sources identified (E-E-A-T) | Author page + cited sources | Medium | Content |
| Brand mentioned consistently across the web | List of the 10 main citations | High | SEO lead |
| AI bot tracking installed | First tracked bot visit | Low | Technical |
| Baseline measurement recorded before any change | Dated response archive | Low | SEO lead |
Download the printable checklist (PDF)
Questions and answers
Did Google's May 2026 guide end GEO and AEO as separate disciplines?
Google's position is that, from Google Search's perspective, optimizing for generative AI is still SEO. The position is internally consistent for Google's own ecosystem (Search, AI Overviews, Discover). It does not extend to ChatGPT, Claude or Perplexity, which run on independent retrieval stacks with different source preferences. GEO and AEO remain meaningful disciplines precisely where Google's perimeter ends.
Should I create an llms.txt file?
Only if your audience includes agents. Four independent studies (Limy, OtterlyAI, ALLMO, SE Ranking) converge on the same finding: llms.txt has essentially no measurable impact on AI Search citations today. The genuine use case is the agentic web: products consumed by Cursor, Claude Code, Copilot or MCP-aware tooling. If you publish a developer-facing API or docs site, ship it. If your audience is end users searching on ChatGPT, your time is better spent on content actions #6 to #10.
How long before this checklist produces measurable results?
Technical actions (#1 to #5) ship in one to two weeks and produce no immediate visibility lift, just the conditions for one. Content actions (#6 to #10) generally start showing up in citations within four to eight weeks. Authority signals (#11 to #15) compound over three to six months. Multi-LLM actions (#16 to #20) depend on the engine: Perplexity updates fastest, Gemini is slower. The realistic horizon to see a Share of Voice trend reverse is one quarter.
Which AI assistant should I prioritize in 2026?
Similarweb's March 2026 data put ChatGPT at 56.7% of AI assistant traffic, down from 86.7% in January 2025. Gemini is around 18%, the remainder distributed across Claude, Perplexity, Copilot and Grok. Concentration is decreasing, not increasing. For most B2B audiences, ChatGPT remains the largest single channel, but pure ChatGPT optimization leaves the other 43% of traffic uncaptured. Track all four major engines.
Can I measure AI visibility without a dedicated tool?
For the first 20 to 30 strategic prompts, yes. Run each prompt weekly on ChatGPT, Claude, Gemini and Perplexity. Log whether you are cited, where in the answer, with what surrounding context. Track referral traffic from chatgpt.com, perplexity.ai, claude.ai and gemini.google.com in your analytics. Past 30 prompts and four engines, the manual cost exceeds the cost of a dedicated tracker. The methodology is what matters: keep it stable across weeks so the data is comparable.
Agencies, consultants and in-house teams: compare the answers generated for your own scenarios, languages and selected models. Our infrastructure, your scenarios, your expertise.