Answer Engine Optimization: what it is, why now, and where to start

Short answer: AEO is the engineering work of getting your brand cited in LLM-generated answers (ChatGPT, Claude, Perplexity, Google AI Overviews). Three things measure: citation rate, citation position, time-to-first-citation. The first three moves are a schema audit and remediation, a 30-line /llms.txt, and one well-structured pillar article per quarter plus a weekly citation poll. Total infra cost under €30/mo on a small Hetzner box.

Most AEO content right now is a pitch deck wearing an article’s clothes. This is the version without the deck. The audience is the engineering lead at a SaaS company who keeps hearing “we need an AEO strategy” from marketing and wants to know if there’s anything real underneath, what’s measurable, and what the first three moves actually look like.

What actually changed

Search used to be a list of links. Between mid-2024 and early 2026 it became answers with citations. ChatGPT Search shipped in late 2024. Perplexity went from a curiosity to a default for a measurable slice of B2B research traffic. Google AI Overviews ate a chunk of zero-click SERP real estate. Claude added web search. By 2026, asking a question and getting an answer with three to five linked sources is what most researchers do before they touch traditional Google.

For a SaaS, the practical implication is narrow and specific. When a prospect asks “what are the leading observability platforms for serverless workloads in 2026”, an LLM no longer hands back a list of SERPs. It hands back four to six brand names with one-sentence summaries and links. If your brand is not in that response, you are not in the consideration set, regardless of how good your SEO ranking is on the underlying queries.

This is not a death-of-SEO story. SEO still drives the bulk of organic acquisition for most B2B SaaS in 2026. AEO is the new layer on top: small in absolute traffic, disproportionate in conversion intent (researchers pre-qualifying vendors), and structurally underbuilt by most companies because the surface is new.

What is actually measurable

Three things measure. Everything else is vanity.

Citation rate. For a fixed list of 30-100 category-relevant queries, what fraction of LLM answers across providers (ChatGPT, Claude, Perplexity, Google AI Overviews) mention your brand? You poll this on a schedule — daily for high-priority queries, weekly for long-tail — and chart the rate over time. A baseline citation rate of 10-20 percent on category-relevant queries in your segment is realistic for a brand starting from zero schema and zero llms.txt infrastructure. The Profound dashboards report higher numbers but those are weighted toward incumbent brands; new entrants start at single digits.

Citation position. When cited, are you the first source mentioned, the third, or buried under “see also”? Perplexity exposes ordered sources directly in the API. ChatGPT and Claude require parsing the generated answer text and matching against the source list. First- position citations correlate with higher click-through. Position is where Profound and similar tools spend most of their dashboard real estate, because in incumbent-vs-incumbent comparisons it’s the more sensitive metric.

Time-to-first-citation. From the moment a piece of content is published, how long before any LLM picks it up? Typical numbers in 2026: 3 to 30 days for Perplexity (live retrieval, fast index), 30 to 90 days for ChatGPT and Claude where retrieval is gated by training cutoffs and crawler refresh cycles. The 30-90 day lag is why “we shipped one article last Tuesday and want to see citations this week” is the most common misalignment between marketing and the work.

What does not measure: keyword density on a page, “AEO score” numbers from tools that grade pages without any actual citation data, vanity metrics like LLM-referrer traffic when no deterministic referrer header exists.

The state of the AEO vendor landscape in 2026

Three categories, in rough order of company size.

Profound dominates the dashboard tier in B2B SaaS. They built the category, they have the integrations, they have most of the named logos. If you are a Series B-plus SaaS with budget for a citation- monitoring platform contract, Profound is what gets bought. Their strength is dashboards and historical baseline; their weakness is that the dashboard is the product — it tells you the score, it does not change the score.

Marketing agencies that added an AEO line. A handful of large SEO agencies repositioned in 2025 and now sell “AEO strategy” alongside content production. Quality varies. The good ones have figured out that schema markup, llms.txt, and structured Q&A content are 80% of the work and they do it well. The bad ones rebranded backlink campaigns as “citation building” and produce articles that no LLM will quote.

Specialized boutiques. A small number of independent operators (Boring Technologies sits here) who do the engineering work — schema audits, JSON-LD remediation, llms.txt drafting, citation-monitoring infrastructure — at the engagement size that does not warrant a Profound contract. Receipts vary; ask for them.

The structural reason a SaaS engineering lead might prefer a boutique is that AEO is mostly engineering work. Schema is a JSON-LD typing exercise. llms.txt is a content-architecture exercise. Citation monitoring is a polling-and-storage exercise. The work resembles SRE or platform engineering more than it resembles content marketing, and a marketing agency that does not have engineers on staff will outsource the technical part anyway.

The first three moves

If you have not done AEO before, here is the order. Each move builds on the previous and each has a measurable test.

Move 1: Schema audit and remediation

Goal: every page worth citing has correct, minimal JSON-LD.

In practice this means:

Article pages emit Article JSON-LD with headline, description, datePublished, author (as a Person), mainEntityOfPage, and inLanguage. Nothing else.
Service pages emit Service JSON-LD with serviceType and provider pointing at an Organization.
Every page deeper than / emits a BreadcrumbList.
The homepage emits Organization and (if applicable) one of ProfessionalService, LocalBusiness, or Corporation.
FAQ-shaped sections emit FAQPage with at least three real questions; if there are fewer than three, do not emit it.

The minimum-viable Article shape on a content page looks like this:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Answer Engine Optimization: what it is, why now, and where to start",
  "datePublished": "2026-05-18",
  "author": {
    "@type": "Person",
    "name": "Tudor Constantin",
    "url": "https://boringtechnologies.com/about"
  },
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://boringtechnologies.com/articles/aeo-introduction"
  },
  "inLanguage": "en-US"
}

That is the full surface for a typical article. Field-by-field explanations are in the schema deep-dive linked at the bottom of this article.

What the audit looks like as work: 3 to 8 hours for a typical B2B SaaS marketing site of 50-200 pages. The output is a JSON-LD diff per template plus a smoke-test that runs schema.org’s validator against every URL in the sitemap.

Test: after deployment, the schema.org validator reports zero errors on every URL. Then wait 7 to 14 days and poll Perplexity for a category-relevant query. If your brand starts appearing where it did not before, the work shipped.

Move 2: llms.txt and content architecture

Goal: a single file at /llms.txt that lists, in machine-readable form, the canonical entry points to your site and a one-sentence description of each.

The convention is small and fast to implement. A 30-line llms.txt that names your services, your pillar articles, and your contact page is enough. The point is not to give LLMs a complete map of your site — they have your sitemap for that. The point is to give LLMs the canonical names and descriptions that you want them to quote when summarizing your brand.

What the work looks like: 1 to 2 hours of writing if your brand positioning is already crisp; 4 to 8 hours if the writing forces you to confront the question of what your brand actually does clearly.

Test: ask ChatGPT or Claude to summarize your company in one paragraph. If the summary uses your llms.txt phrasing, the file is doing its job.

Move 3: One pillar article per quarter, and the citation poll

Goal: one substantive article (1500-3000 words) per quarter that answers a category-defining question your prospects ask, plus a weekly poll of 30-50 category queries to track whether the article moves citation rate.

The article needs to be three things: factually correct, well- structured (clear sections, code or examples where relevant, a real conclusion), and discoverable (the schema and llms.txt from moves 1 and 2 do this part). It does not need to be SEO-stuffed or backlink-padded. LLMs are weighted toward content that answers the question. They do not weight backlinks the way classical search did.

The poll is a small piece of infrastructure: a Postgres table, a cron job, an API call to each LLM provider, a row inserted per provider per query per day. Total monthly cost on a small Hetzner box: under €30 including the API budget for daily polls on 50 queries. If that sounds like more engineering than your marketing team owns, that is exactly the gap a specialized AEO vendor fills.

Test: 30 days after the pillar article ships, the citation rate on the queries it targets has moved up at least one provider’s worth (typically Perplexity first, then Google AI Overviews, then ChatGPT, then Claude — in roughly that lag order).

How we use AEO on our own sites

Boring Technologies works as the agency layer on Cristina Constantin, a Romanian civil-law cabinet, used as the living lab for the AEO methodology. The same playbook — schema, llms.txt, weekly poll, one pillar article per cadence — is what clients buy. The work is visible: the source on every Cristina page shows the JSON-LD; the /llms.txt is public. First Perplexity citations on RO succesiuni queries showed up inside 30 days of publishing. That is the timeline to set expectations against.

What this article is not

This is not a list of twenty AEO tactics ranked by promised visibility multiplier. The work is small in number of moves and large in care taken per move. Most of the lift comes from doing the schema correctly the first time, writing one strong article instead of ten thin ones, and running the poll long enough to see the trend.