What is Answer Engine Optimization (AEO)?
Answer engine optimization (AEO) is the work of making your business legible to the models that answer buyers’ questions. When someone asks ChatGPT “who’s the best HVAC company in Dallas,” the answer the model gives is already the shortlist. The buyer never sees a Google results page. They see a name, a sentence, maybe a link. If your business isn’t in that sentence, you weren’t considered.
AEO is how you become one of the names that gets considered. This page covers what it is, what it isn’t, the four-layer stack that makes a business legible to a model, why the work is urgent now, and where to start.
AEO is not SEO with a new label
SEO optimizes for a search index. The target is a ranking on a results page a human will scroll through.
AEO optimizes for a model. The target is a quotable sentence a model will lift from your site when it writes the answer.
The two overlap at the technical floor, since a page that can’t be crawled fails both, but the ceiling is different. A page can rank in Google and still be invisible to Claude, because Google will index JavaScript-rendered content that Claude won’t execute. A page can be first in the map pack and still be missing from Perplexity’s answer, because the information Perplexity needed to quote isn’t actually on the page.
Three concrete differences:
- The reader is a model, not a human. Models don’t scroll, don’t click, and (today) mostly don’t run JavaScript. They read server-rendered HTML and structured data. If your copy only appears after a client-side render, the model sees an empty page.
- The target is a sentence, not a session. SEO tries to earn a click. AEO tries to earn a quote. You don’t need the buyer to visit, you need the model to carry a specific sentence of yours into its answer.
- The signals are measurable in a new way. Backlinks and keyword density don’t go away, but they stop being the whole story. Robots access, schema markup, named presence in trusted sources, and specific answers on-page start to dominate.
What AEO is not
- It is not “write content for AI.” Models are not an audience you flatter. They are a crawler that happens to summarize. The work is the same work good product writing has always been: be specific, be answerable, be true. But it has to live in HTML a model can read.
- It is not a prompt-engineering trick. You cannot prompt your way into an answer engine’s output from the outside. The model’s answer is built from what the model has read on the open web, under its own crawl.
- It is not a replacement for a working website. If your site is broken, slow, or missing basic information, AEO fixes none of that. It stacks on top of the fundamentals; it does not substitute for them.
The AEO stack
Four layers, bottom-up. If any layer is broken, the layers above it don’t matter.
1. Access. robots.txt has to let the right crawlers in: GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot (Perplexity), Google-Extended (Google’s AI training and answer crawl), and Applebot-Extended (Apple Intelligence). A business that silently disallows these is choosing to be invisible to the systems the shortlist is built from. For almost every local service business, the right call is: allow them.
2. Legibility. The page a crawler fetches has to actually contain the business’s information in HTML. Server-rendered or statically generated. Core text, services, service areas, hours, phone, address, present in the DOM at response time, not injected by a React tree on the client. Images with real alt text. JSON-LD structured data for LocalBusiness, Service, and FAQPage where it fits.
3. Specific, lift-able answers. This is the layer most sites fail. Marketing copy that reads well to a human (“we’re committed to excellence”) gives a model nothing to quote. The model needs sentences that can be pulled whole into an answer. “We respond to new HVAC emergency calls within 90 minutes anywhere inside Loop 820 in Fort Worth” can be an answer. “We care about our customers” cannot.
4. Named presence in trusted sources. Models answer with names they have seen before, in places they trust: local directories the search index has indexed for a decade, review sites with real review volume, local-paper mentions, Chamber-of-Commerce listings, and trade-association rosters. The link graph still matters; what shifted is that the model is now a reader of the link graph, not just a search algorithm.
Why this is urgent, not optional
Answer-engine share of discovery is growing month over month. The buyers using it skew exactly the way local service businesses care about: higher intent, less patient, more likely to go with the first name they’re given. A buyer who asks ChatGPT for a roofer and gets three names has already shortlisted. The follow-up search is a verification, not a discovery.
Businesses that treat AEO as “future problem” are letting that shortlist get fixed without them in it. Businesses that treat it as a weekly discipline (audit, fix, measure, repeat) are becoming the names the model learned.
NoCode Labs measures this with a Digital Presence Score. The pillar at /what-is-dps walks through what moves it. The short version: the signals above, scored each week, with the deltas that matter surfaced so an operator can approve the next fix.
Where to start
- Open your
robots.txt. If it blocks GPTBot, ClaudeBot, PerplexityBot, or Google-Extended, unblock them. - View-source any page you consider important. If the body copy isn’t in the HTML before JavaScript runs, that page is invisible to models.
- Rewrite your top three answer pages (“what do you do,” “where do you do it,” “what does it cost”) into sentences specific enough to quote.
- Audit your presence in the directories and review sites that already rank for your category. The model trusts them because the search index does.
- Then measure. Weekly. The score going up is the only proof that the fix stuck.
Frequently asked questions
See how we measure it.
Digital Presence Score is the weekly number built from the signals above. We publish our own, and the breakdown, so you can judge the methodology before you trust the number.