What is Answer Engine Optimisation (AEO)?

AEO is the practice of structuring your online presence so AI assistants — ChatGPT, Perplexity, Google AI Overviews — cite your business when answering relevant queries. Unlike SEO, AI engines select sources, not pages.

How does Citura calculate my AI visibility score?

Citura analyses 6 dimensions: robots.txt/llms.txt configuration, structured data (schema.org), answer-first content depth, agent.json presence, and active citation signals. Each dimension is weighted; the aggregate is your 0–100 score.

How long before I see results with Citura?

AI citation indexes update slowly — expect meaningful movement in 4–8 weeks. Your score improves within days as optimisations are deployed; actual citations follow as AI engines re-crawl your domain.

Does Citura guarantee AI citations?

No. Citura sells the system and the signals — not a placement. Anyone promising "rank #1 in ChatGPT" is misleading you. We make your business structurally citable; the AI engine decides when to cite.

Is my data safe with Citura?

Citura only analyses publicly accessible data on your domain. No credentials are required. All processing runs on EU infrastructure (Google Cloud Run, region europe-west6).

Do AI engines cite social media posts?

Rarely and inconsistently. Most social platforms block AI crawlers or have rate-limited access agreements. Perplexity cites Reddit threads frequently (Reddit has a data licensing agreement with Google and distributes data to several AI companies). LinkedIn posts, X/Twitter, and Facebook are not reliably cited because their crawl access is restricted. Your primary citation surface should be your own domain.

Does getting cited once mean I will always be cited?

No. AI citation is query-specific and varies with each crawl cycle. A page cited today for query A may be replaced next week by a fresher or better-structured page. This is why citation monitoring — tracking your frequency across target queries over time — is more informative than a single citation event. Trends matter more than individual occurrences.

Can I see when an AI bot crawls my site?

Yes, through your server access logs. GPTBot, PerplexityBot, ClaudeBot, and Google-Extended all have documented user agents. Filter your access logs for these user agents to see crawl frequency and which pages they visit. Low crawl frequency (less than weekly for important pages) suggests your domain authority or internal linking needs improvement.

Do AI engines cite Wikipedia?

Yes, frequently. Wikipedia is heavily represented in AI training data and is accessible to all AI crawlers. Having your business mentioned in a relevant Wikipedia article (as an example, reference, or in a list) creates a persistent citation signal across all AI systems. This is one of the highest-leverage GEO tactics available, though it requires genuine notability to be accepted by Wikipedia editors.

How AI Assistants Decide Which Sources to Cite

ChatGPT, Perplexity, and Google AIO each use different source-selection mechanisms. This guide explains the observable logic behind each system and what signals increase your probability of being selected as a cited source.

AI answer engines select sources through a multi-stage process: query → retrieval → re-ranking → excerpt extraction → citation. Each stage filters candidate pages. Your business can only be cited if it passes all stages. The signals that determine passage differ meaningfully between ChatGPT, Perplexity, and Google AIO.

The general architecture: retrieval-augmented generation

Modern AI answer engines use Retrieval-Augmented Generation (RAG). When a user submits a query: (1) a search component retrieves candidate documents; (2) a re-ranking model scores the candidates for relevance and quality; (3) the language model reads the top candidates and generates a synthesized answer; (4) source URLs are attached to claims in the answer. The critical insight is that selection happens at step 2 — a page that is retrieved but poorly structured will lose to a page that is both retrieved and densely informative.

ChatGPT (with web search)

ChatGPT's web browsing uses a Bing-powered search API to retrieve candidates, then applies a re-ranking model to select the best passages. Observed selection factors:

GPTBot access — pages blocked to GPTBot in robots.txt are excluded entirely. This is the most common and easily fixed barrier.
Answer-first content structure — ChatGPT excerpts the first complete answer it finds. Pages that open with the direct answer are cited with the correct excerpt; pages that bury answers produce low-quality excerpts.
Bing index presence — ChatGPT's web search draws from Bing's index. Pages not indexed by Bing are invisible. Submit via Bing Webmaster Tools if your Bing indexation is low.
JSON-LD schema — FAQPage and Article schema improve passage identification quality.
Domain freshness signals — recently updated pages (reflected in dateModified in Article schema) are preferred for time-sensitive queries.

Perplexity

Perplexity operates its own web crawler (PerplexityBot) and builds a proprietary index. Its re-ranking is notably citation-aggressive — it frequently cites more sources per answer than ChatGPT and tends to pull from longer-form, structured content. Key signals:

PerplexityBot access — like GPTBot, must be allowed in robots.txt. Perplexity also respects llms.txt and agents.json for context.
Subheading density — Perplexity's extraction model pulls from H2/H3 sections specifically. Pages with clear subheadings matching query intent are cited at the section level, not just the page level.
Numerical specificity — Perplexity strongly prefers pages with specific data (percentages, dates, named entities). Vague qualitative claims are replaced by pages with numbers.
Source diversity signals — Perplexity appears to avoid over-citing a single domain and deliberately includes diverse source types (news, official docs, expert blogs).

Google AI Overviews (AIO)

Google AIO is the most complex system because it integrates directly with Google's existing search infrastructure. Source selection combines traditional PageRank signals with AI-specific quality factors:

E-E-A-T signals — Experience, Expertise, Authoritativeness, Trustworthiness. AIO heavily weights pages from demonstrable experts (author credentials, organization affiliation, consistent topical coverage).
Featured snippet eligibility — Pages that already appear in featured snippets for a query are disproportionately selected for AIO. Featured snippet optimization (direct answers, table formatting, numbered lists) directly feeds AIO selection.
Freshness — For time-sensitive queries, AIO prefers pages updated within the last 90 days.
Google-Extended bot access — Google has a separate crawler (Google-Extended) for AIO training. Some sites block it; doing so reduces AIO citation probability.

What all three systems agree on

Despite their differences, all three AI answer engines consistently select pages that: (1) provide the direct answer to the query in the first paragraph; (2) use clear H2/H3 semantic structure; (3) are accessible to their respective crawlers; (4) have valid structured data; (5) are from domains with external references (backlinks, directory listings, citations in other content). These five signals represent the non-negotiable baseline for AI citation eligibility across all major systems.

What gets pages excluded

Pages are actively filtered when they: block the AI crawler in robots.txt; require authentication or JavaScript to render (many SPAs are invisible to AI crawlers); contain thin content under ~300 words; have no external links pointing to them; or are on domains with no AI-accessible crawl history. Paywalled content is also excluded — AI systems will not cite behind-login pages.