The 30-second answer
AEO in one paragraph.
AEO (Answer Engine Optimization) is the work of making a website easy for AI engines to read, summarize, and cite. The output of SEO is a ranked list of blue links. The output of AEO is your brand named inside the answer itself, often without the user ever clicking through. A site that is good at AEO has clean structured data, a named author behind every page, an llms.txt file at the root telling engines what the site is about, FAQ blocks marked up so questions and answers are discrete, and substantive content that reads like a reference rather than a brochure.
If your buyers are typing questions into ChatGPT, Perplexity, or Google AI Overviews, AEO determines whether your brand shows up in the answer.
Search behavior is changing faster than most brands realize.
For two decades, the way to be found online was simple: rank on Google, get clicks, convert traffic. Search engines returned a list of ten blue links, the user picked one, the site they picked got the visit. The whole industry of SEO was built around moving up that list.
That model is breaking. Google's AI Overviews now appear on roughly 30 percent of US commercial searches and growing. ChatGPT crossed 200 million weekly active users. Perplexity has become a primary research tool for B2B buyers. Claude is increasingly used inside enterprise workflows. In every one of these surfaces, the user no longer sees a list of links to choose from. They see an answer, often with two or three brand names cited inline as sources.
of US commercial searches now end with the user reading an AI-generated answer instead of clicking a link. There is no second page of results to optimize for. You are either cited or invisible.
If your brand is one of the cited sources, the user reads about you in an authoritative voice and arrives at your site already half-converted. If your brand is not, you are invisible. This is the gap AEO fills. The mechanics are different from traditional SEO because the engines parsing your content are different (large language models, not crawler-and-index pipelines), and because the success metric is different (cited in an answer, not clicked through to a page). The good news is that the underlying signals overlap heavily, so a site investing in AEO usually picks up SEO improvements as a side effect.
Three layers AI engines evaluate before citing a source.
A useful way to think about AEO is in three stacked layers. Each one is necessary. A site weak on any of them gets passed over for whichever competitor is stronger on all three.
Machine readability
Can the engine parse the page cleanly and understand what it is? A site that fails this layer is invisible regardless of how good the underlying content is. The engine never gets far enough to evaluate it.
- Valid JSON-LD schema
- llms.txt at root
- Clean sitemap
- AI crawlers allowed
- Semantic HTML
Citability signals
Once the engine can read the page, does it look like a credible source to cite? Citability is signaled by attribution, structure, and concrete claims that are quotable rather than vague marketing prose.
- Named author
- FAQPage schema
- Quotable claims
- Publication dates
- Entity coverage
Authority
Citability is necessary but not sufficient. Authority is what tips the engine toward citing one source over another when both are equally readable. It compounds slowly, which is why early AEO investment pays off later.
- Quality backlinks
- Brand mentions
- Domain age
- Third-party reviews
- Industry coverage
How AI engines actually decide who to cite.
Each engine works slightly differently, but the patterns are converging. Based on what we see in our own monitoring across client AEO engagements, here is what consistently shows up in the citation set.
The query gets decomposed into discrete claims.
If a user asks "best calgary marketing agency for ecommerce brands", the engine doesn't search for that exact string. It decomposes the query into discrete sub-claims (what makes a marketing agency good, what does ecommerce work specifically need, who serves Calgary, what does the user actually want to evaluate) and looks for sources that answer each sub-claim with a quotable line.
Sources are ranked on parseability first, authority second.
A perfectly authoritative source with messy structured data often loses to a cleanly parsed source with moderate authority. This is the inversion that catches established brands off guard. If your site is hard for the engine to read, it doesn't matter how strong your domain is. The engine cites the cleaner source.
Citations cluster around recency and consensus.
Engines prefer sources that have been recently updated and that agree with the broader web's consensus on the topic. Outlier claims need extra-strong evidence to be cited. Recently dated content with proper Article schema gets weighted more than undated marketing pages on the same topic.
FAQ blocks are lifted directly into answers.
If a page has FAQPage schema with a question that matches the user's query, the engine often lifts the answer text into its response verbatim. This is the highest-leverage AEO move available, which is why FAQ schema is one of the first things we ship on any new engagement and why the audit tool weights it heavily.
Brand entity coverage matters more than keyword density.
An AI engine does not count keyword frequency the way a 2010 SEO crawler did. It evaluates whether the brand has consistent entity coverage (Wikipedia, Crunchbase, Knowledge Panel, third-party reviews, industry coverage) and whether the source page is recognized as a legitimate publisher in its category. Spamming a target keyword across the page does nothing. Building entity recognition does.
The two are not in conflict, but they are not the same.
The most common mistake we see is treating AEO and SEO as alternatives. They are not. They share infrastructure but optimize for different outcomes. We cover the full breakdown in a separate post; here is the short version.
SEO
Optimizes for being clicked.
AEO
Optimizes for being cited.
The infrastructure overlap (clean technical health, substantive content, structured data, authority) means investing in one usually improves the other. The divergence is where the leverage is. SEO does not require llms.txt or FAQ schema, but AEO leans on them heavily. AEO does not require deep keyword targeting at the URL level, but SEO does.
If your buyers are still primarily using Google's classic search, SEO carries more weight. If they are increasingly using ChatGPT, Perplexity, or AI Overviews, AEO is the place to invest. For most brands today, the answer is "both, with AEO weight increasing each quarter."
Each engine cites differently. Here is what matters per engine.
The four engines that matter today (ChatGPT, Perplexity, Google AI Overviews, Claude) work on different infrastructure, refresh on different cadences, and weight signals differently. Optimizing as if they are interchangeable underperforms. The shape of the work is the same, but the priorities shift per engine.
| Engine | Refresh cadence | Citation style | Over-weights | Penalty traps |
|---|---|---|---|---|
Perplexity |
Near real-time | Inline numbered citations + sources sidebar | Recency, structured data, clean schema | Stale content with old dateModified |
Google AI Overviews |
Continuous (Google index) | Inline citations with source pill, expand-to-list | E-E-A-T signals, schema, traditional SEO authority | Thin content, missing author attribution |
ChatGPT |
Quarterly training + live RAG | Sources panel at end, footnoted citations | Brand authority, broad web mentions, llms.txt | Blocked GPTBot, no entity coverage outside own site |
Claude |
Periodic training + connector RAG | Inline links, attribution near claim | Quotable claims, named authors, factual density | Marketing prose without specific claims to cite |
The practical takeaway: clean schema and llms.txt cover the foundation across all four. Beyond that, Perplexity rewards a recent dateModified, ChatGPT rewards entity coverage outside your own site, Google AI Overviews rewards traditional E-E-A-T (named author, citations, brand mentions), and Claude rewards concrete quotable claims over marketing prose. Same building blocks, slightly different proportions.
What this looks like in practice.
AEO is easy to describe and harder to demonstrate. Here is the citation curve we typically see across the first six months of an engagement, broken down by engine. The shape repeats: Perplexity moves first, Google AI Overviews follows within weeks, ChatGPT and Claude lag while their training and RAG layers catch up.
Representative DTC engagement · monthly cited mentions per engine
Citations per engine, month 6
monthly cited mentionsPerplexity moves first.
Citations on Perplexity start within weeks of clean structured data shipping. It is the fastest signal that AEO is working before the slower engines catch up.
Citations compound.
Month 0 to month 1 shows small numbers. Month 3 to month 6 shows the curve bending up. The engines start trusting a source as it becomes a recurring citation.
ChatGPT and Claude follow.
The slower-refreshing engines lag, then catch up once the entity coverage and authority signals reach their training cycle. By month six they are citing too.
The six things every site should ship first.
Before any sophisticated content strategy, every site needs a baseline. These are the lowest-cost, highest-leverage AEO moves. Most can be done by a competent developer in a single sprint. The next section gives you the actual code to copy.
Add llms.txt at the site root
A plain text file telling AI engines what your site is about and which pages matter most. Critical signal, costs nothing. See the full guide.
Allow AI crawlers in robots.txt
Explicitly allow GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot. If you blocked them previously to "protect content," you are blocking yourself out of citations.
Ship Organization JSON-LD on every page
A small block of structured data identifying your brand. Lets engines link your pages to a single entity. Five minutes of work, sitewide impact.
Add FAQPage schema where you have Q and A content
FAQ blocks are lifted directly into AI answers. Schema must match visible Q and A on the page. Highest-impact single move on most sites.
Name the author on every editorial page
Even a single meta name="author" tag flips the E-E-A-T signal. Person schema with bio and credentials is the stronger version when the author is real.
Audit and iterate
Once the baseline is shipped, measure where you stand and where the gaps are. Then iterate monthly until you are reliably cited in your category.
Try it now
See where your site stands.
Type a URL. We will fetch the page, run the same twelve AEO checks our audit tool runs, and show you a real score in about ten seconds. No email required.
What not to do, because doing it makes things worse.
The cost of bad AEO is higher than the cost of no AEO. Engines penalize sites that look like they are trying to game the system. These are the most common ways brands shoot themselves in the foot.
Faking FAQ schema with no visible Q&A.
Adding FAQPage JSON-LD to a page that has no visible questions and answers is schema spam. Google has explicitly demoted pages caught doing this. The schema must mirror the visible content exactly.
Blocking AI crawlers in robots.txt.
Some agencies recommend blocking GPTBot and ClaudeBot to "protect" content. This is throwing away the citation channel entirely. The user reads the answer somewhere; it just won't be from your site.
Walls of marketing prose with no quotable claims.
"We are a leading provider of innovative solutions" is unciteable. AI engines need discrete, concrete claims to lift. Pricing, timelines, processes, results, named methodologies all work. Adjective stacks do not.
Hiding answers inside accordions only.
Accordions are fine for design, but make sure the answer text is in the DOM at page load (not lazy-loaded by JS on click). If the engine has to execute JavaScript to read your FAQ answer, most engines will skip it.
Stuffing keyword density.
Repeating "calgary marketing agency" 30 times on a page does nothing for AI engines and triggers spam classifiers in Google. Engines look for entity coverage and topical depth, not raw frequency.
Skipping author attribution because "we are a brand."
Even an Organization can be the named author. The single biggest E-E-A-T signal is a meta name=author tag, which takes ten seconds to add. Skipping it is leaving free points on the table for every page on your site.
Frequently asked.
AEO stands for Answer Engine Optimization. It is the practice of structuring a website so that AI engines like ChatGPT, Perplexity, Google AI Overviews, and Claude reliably cite it when generating answers to user questions.
Yes. SEO optimizes for ranking in a list of links, where the user clicks through to the site. AEO optimizes for being cited inside an AI-generated answer, where the user reads the answer and may never click. They share infrastructure (technical health, content quality, authority) but the success metrics are different.
No. The two work in parallel. Most of the underlying signals overlap (site health, content depth, structured data, backlinks, brand authority), so investing in AEO improves SEO and vice versa. Brands that abandon SEO to chase AEO usually lose both.
AI engines weight a small set of signals heavily: structured data (especially Organization, Article, and FAQPage schema), the presence of an llms.txt file, named authors and entity coverage, content that directly answers a question with a clear claim, and traditional authority signals like backlinks and brand mentions. Sites that look like reference material to a parser are cited more often than sites that look like marketing copy.
llms.txt is a plain text file at the root of a site (like robots.txt) that tells AI engines what the site is about, which pages to prioritize, and how to interpret the content. It was proposed in late 2024 by Jeremy Howard and has been adopted by AI engines as an authoritative signal. Sites with a clean llms.txt are easier for AI engines to summarize and cite.
FAQPage is a JSON-LD schema type that marks up question and answer pairs on a page so engines can lift them directly into AI-generated answers. It is one of the highest-impact AEO signals because the answer is already formatted as a discrete claim, which is exactly what an AI engine needs to cite. The questions in the schema must match the visible questions on the page.
Months, not weeks. Each AI engine refreshes its index on a different cycle. Perplexity tends to be fastest, Google AI Overviews next, ChatGPT and Claude lag because their training-data refreshes are less frequent. Expect first measurable citations within 60 to 90 days of a clean implementation, meaningful share of voice within six months.
Yes for the foundational layer. The basics (llms.txt, Organization and FAQPage schema, named author meta tags, sitemap, allowing AI crawlers in robots.txt) are technical work any competent developer can do in a day. Where agencies add value is in ongoing content strategy, citation tracking, structured data depth, and the iterative tuning that turns a baseline into market-leading citation share.
Make sure you're chosen
Audit your site, then build the plan.
Two minutes to see where you stand. Then either ship the baseline yourself, or talk to us about doing it for you.