AEO and GEO for AI-built apps: getting cited by ChatGPT and Perplexity
When someone asks ChatGPT 'what's the best compliance scanner for Lovable apps?' you want your answer to be the one it cites. This is how AEO and GEO actually work — practical tactics evidence-based on what gets cited, not marketing fluff.
AEO vs GEO vs SEO in one paragraph
SEO = ranking on Google's traditional ten blue links. AEO (Answer Engine Optimization) = being the source for direct answers in Google's AI Overview, ChatGPT Search, and Perplexity. GEO (Generative Engine Optimization) = being cited by LLMs anywhere they generate responses sourced from the web. The three overlap heavily — most things you do for one help the others. The differences matter at the margins.
Perplexity drives roughly 4–8% of search traffic for tech queries (mid-2026). ChatGPT's web-browsing feature drives another ~3%. Google's AI Overview shows above traditional results for ~20% of queries. If your content isn't structured for AI engines to cite, you're invisible on that growing share of traffic.
The 7 things that actually move citations
1. FAQ blocks on every page that can have them
AI engines love FAQ schema. A page with 6 Q&As is essentially 6 separate citation surfaces. Each question should be a real query someone might type into ChatGPT. Each answer should be 40–80 words — long enough to be substantive, short enough to lift as a direct quote in an AI response.
Question phrased as a real query ('Does Comply Code work with Lovable apps?') · Answer leads with yes/no/the-direct-answer · Then 1–2 sentences of reasoning · Optional: link to a deeper article. Wrap in FAQPage JSON-LD.
2. Specific data points and dates
'A lot of people get sued' doesn't cite. '3,117 ADA Title III lawsuits filed in 2025, +37% YoY' cites. LLMs preferentially pull content that has concrete numbers, dates, and named sources because hallucination-penalty training favors specificity. Every important claim in your articles should have a number, a date, and a citation behind it.
3. Citations linked to primary sources
When you make a factual claim, link to the actual primary source — court filing, regulation, academic paper, official documentation. AI engines do reverse-citation checks: a page that links to five authoritative sources gets weighted higher than the same content with no sources. The citation links don't need to be visited by the user; they need to exist for the crawler.
4. Lists and tables over prose
LLMs parse structured data more reliably than prose. If you have five patterns of something, write them as a list with one-sentence descriptions. If you have a comparison, use a table. The Aggarwal et al. paper on GEO (arXiv 2311.09735, 2023) found that lists improved citation rates by 30–40% compared to the same content delivered as paragraphs.
5. JSON-LD structured data, varied by page type
Beyond Organization and WebSite, add specific schemas where they fit. Each one is a separate signal to LLMs about what kind of content this is and how to cite it.
- FAQPage on any page with 3+ Q&As — the highest-impact addition.
- Article on blog posts and explainers, with datePublished and author.
- HowTo on tutorial pages, with numbered steps and estimated time.
- SoftwareApplication on product or demo pages, with offers.
- DefinedTermSet on glossary pages, with hasDefinedTerm[] entries.
- BreadcrumbList on interior pages, defining the navigation hierarchy.
- Dataset on data-citation pages, with measurement details.
6. Visible dates and explicit authorship
Every article needs a visible date (not just buried in metadata) and an explicit author — organization or named person. LLMs aggressively discount content without dates because they can't tell if it's current. A 'Last updated 2026-05-14' line visible at the top of the page increases citation likelihood significantly versus the same content with no date.
7. Welcome AI crawlers in robots.txt
Most sites block AI crawlers (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended) by default, sometimes accidentally via a template. If you want to be cited, explicitly Allow: / for each. Comply Code does this — see our public robots.txt for the exact block.
User-agent: GPTBot · User-agent: ChatGPT-User · User-agent: ClaudeBot · User-agent: PerplexityBot · User-agent: OAI-SearchBot · User-agent: Google-Extended — for each, Allow: /. These should be separate User-agent blocks, not commas.
What doesn't actually work · skip these
A lot of 'AEO' and 'GEO' advice online is recycled SEO content with new acronyms. Things that don't actually move citation rates:
- Keyword stuffing for 'natural language queries' — LLMs ignore this and may down-weight it.
- 'Optimizing for voice search' — same content, different rebranded acronym.
- Embedded video transcripts unless the video is the main content of the page.
- Generic 'authority' signals like awards or badges — LLMs verify; they don't trust glossy badges.
- Paying for placement in 'AI training datasets' — these companies are selling vapor.
How to know it's working
AEO and GEO are harder to measure than traditional SEO because there's no Search Console for ChatGPT. Three signals to track manually:
- Direct: ask ChatGPT, Perplexity, and Google AI Overview the queries your customers ask. Search for '[your category] for [your audience]' and similar phrasings. See if you appear in the citations. Re-check monthly.
- Referral: check your web analytics for traffic from chat.openai.com, perplexity.ai, claude.ai, and bing.com. The number is small but growing — and it's high-intent traffic.
- Branded: watch for direct traffic to your domain when you haven't run ads — that's brand awareness driven by being cited.
A worked example: this article
This article is structured for AEO and GEO. Look at the patterns:
- Opens with a direct definition ('AEO vs GEO vs SEO in one paragraph').
- Major sections numbered and titled with a clear topic word.
- Specific numbers and dates appear throughout ('3,117 lawsuits in 2025', 'Perplexity drives 4–8% of tech traffic').
- Citations link to primary sources (the GEO paper on arXiv, schema.org).
- FAQs cover the long-tail questions ChatGPT users actually type.
- JSON-LD on this page includes Article + FAQPage schemas, set by the article's data block.
AEO and GEO are emerging fields. The tactics here are evidence-based as of mid-2026, but LLM citation behavior is moving fast. Re-check this article every 3–6 months — what works in summer 2026 may not work the same way in 2027.
Common questions.
What's the difference between AEO and GEO?
They overlap heavily. AEO (Answer Engine Optimization) focuses on getting your content used as the direct answer in tools like Google's AI Overview, ChatGPT Search, and Perplexity. GEO (Generative Engine Optimization) is broader — getting cited anywhere an LLM generates content from web sources. Most tactics serve both; the differences are at the margins of which platforms each term emphasises.
Will SEO become obsolete?
No. Google still drives 80–90% of search traffic for most categories. AEO and GEO are additive — you do both. The good news: most things that help SEO (good content, structured data, fast loading, clear authorship) also help AEO and GEO. The 12-step SEO checklist in our companion article is also the foundation for citation-readiness.
Do AI engines actually cite indie websites?
Yes — that's the opening for indie operators. Unlike Google, which favors established domains, AI engines preferentially cite specific structured content over generic 'authoritative' brand sites. A well-structured 1,500-word piece from your site can out-cite a Forbes article on the same topic. The Aggarwal et al. paper documents this empirically.
Should I block AI crawlers to protect my content from being used in training?
Different question. Training crawlers (GPTBot for OpenAI training, ClaudeBot for Anthropic training) are separate from search crawlers (OAI-SearchBot, ChatGPT-User, Claude-User, PerplexityBot). You can allow search crawlers (to be cited) and block training crawlers (to opt out of training corpora) — each has its own user-agent string in robots.txt. Most indie operators allow both.
What about hallucination risk if LLMs cite me?
LLMs do occasionally misquote sources. The mitigation: make your content unambiguous — specific numbers, direct claims, primary citations. LLMs have lower hallucination rates when copying from structured, specific content versus paraphrasing prose. Vague claims invite vague citations.
How quickly does AEO and GEO show results?
Faster than SEO. LLMs re-ingest content every few days versus Google's weeks-to-months re-crawl. A new article well-structured for AEO can start appearing in AI citations within 1–2 weeks if it covers a topic the LLM regularly gets asked about. That's substantially faster than the 3–6 month Google ranking timeline.