Our AEO methodology: how we make you the answer AI recommends

Answer Engine Optimization (AEO) is a simple goal with a hard execution: be the tool ChatGPT, Claude, and other answer engines name when a buyer asks for the best option in your category. This is the method we use to get there — including the counterintuitive parts we learned the hard way.

The uncomfortable truth: llms.txt won't save you

Every AEO checklist starts with llms.txt, and almost every one overstates it. In practice it is near-useless for citations today: large-scale crawler studies find it barely fetched, Google has publicly said it won't use it, and removing it has been shown to improve some citation-prediction models. We still ship a clean one — it's cheap hygiene and a forward bet — but if that's your whole AEO strategy, you will not get recommended.

What actually drives AI citations

The signals that move share-of-answer for a B2B SaaS are well understood. In rough order of leverage:

• Your own site. This is the one most AEO advice buries, and it is the biggest by a distance: when the question names the brand or its category, 74–78% of AI citations for business software go to the vendor's own pages (Profound, Ranqo, 2026); on open "best X" questions the weight shifts to third-party pages. Answer-first structure, named customer quotations, and content a non-JavaScript crawler can actually read.
• Comparison and "alternative-to-X" pages — answer-first, with a quotable feature table. The ranked "best-of" listicle is among the most-cited content formats in every 2026 dataset we have checked.
• The Princeton content stack — citing sources, adding statistics, and using quotations each lift citations 30–40%; fluent, answer-first writing helps; keyword stuffing actively hurts.
• Community presence — Reddit is the single most-cited domain in AI answers, and LinkedIn is a top B2B source. You can't fake your way in; you have to actually be there.
• Third-party review profiles — G2, Capterra, TrustRadius (US) and B2B Stack, Reclame AQUI (Brazil). Being listed and claimed is what matters here, and it's binary: claim the profile once and it keeps working. We used to rank this first, on a widely-repeated "3× citation multiplier" figure. We measured it ourselves and could not reproduce it — review pages are a low-single-digit share of AI citations. Collect reviews because buyers read them, not because they move the model.
• Original, first-party data — an annual study earns editorial coverage, which is what feeds Wikipedia and Knowledge Panels. It's an un-copyable, durable source of citations.

Measure, then execute — done for you

Diagnosis is the easy half. We run your brand through real, web-grounded answers from ChatGPT and Claude, track a Reputation score, and map the exact citation gap against your competitors. Then — and this is the part most tools skip — our AI agents produce the fixes (comparison pages, structured data, entity claims, review playbooks, publisher outreach), a specialist reviews every change for brand safety, and we ship it to your site via your CMS or a GitHub PR, then re-verify against the live site. You approve; you never write a line.

What we will never do

We never fabricate reviews, testimonials, case studies, or citations. Earned inclusion only. Fake reviews violate G2/Capterra terms, the FTC's 2024 rule, and Brazil's CDC — and they backfire: platforms and the LLMs that read them detect and penalize astroturf. When we help you get reviews, they come from your real, verified customers, in their own words.