By Karthik Kumar · Founder, Cafiyn Innovations

How to rank on Google + AI in 2026.

AEO (Answer Engine Optimization), GEO (Generative Engine Optimization), and SEO (Search Engine Optimization) are converging in 2026. This is the exact playbook that gets a fresh domain indexed and cited across Google AI Overviews, Bing Copilot, ChatGPT search, Perplexity, Claude search, and Gemini within the first 30 days. Cafiyn Pulse ships with every technique in this guide already applied, so this doubles as the working case study.

Try Cafiyn Pulse (free)

Definitions

The three, in plain language.

SEO optimizes for a page's ranking in a list of blue links. AEO optimizes for being the single quoted answer to a specific question. GEO optimizes for being one of the sources a generative AI cites when it synthesises a longer answer across several pages. All three overlap in 2026 and most real work covers them together.

The old SEO framing (rank number one, drive a click) is now the top of a wider funnel. Below it: appear inside a Google AI Overview, be the source of a Perplexity answer, become a cited link in a ChatGPT search response, get quoted verbatim by Claude search, show up in a Gemini snapshot. Every step of that funnel favours the same fundamentals plus a few new signals.

The 12 signals

The signals that actually move the needle in 2026.

Ranked roughly by leverage for a small team in the first 90 days. Every one of these is in production on this site, inspect the source of any page and you will see it.

FAQPage schema

The single most extracted schema for AI answer engines. 5-10 real Q&A pairs per important page, with direct-answer text in acceptedAnswer. Do not stuff.

HowTo schema

Wins "how do I X" queries. Include name + text per step. Google renders as an expandable how-to card; LLMs love the numbered structure.

BreadcrumbList

Replaces raw URLs in search results with readable breadcrumb text. Small win visually, medium win for click-through.

SoftwareApplication / Product / Article

Type-specific schemas trigger rich cards (rating, price, author). Pick the closest type per page and fill in every relevant field.

Speakable

CSS selectors on Article schema that tell Google Assistant which sections are safe to read aloud. Free, future-facing for voice.

Person + Organization (author, publisher)

Consistent Person schema on every content page, cross-linked to a persistent Organization. Helps LLMs deduplicate you as an entity.

IndexNow

One POST to indexnow.org and Bing, Yandex, DuckDuckGo re-crawl instantly. Bing feeds three AI engines. GitHub Action costs nothing to run.

llms.txt

Curated table-of-contents file at your site root for LLM crawlers. Emerging standard; low cost, growing influence.

Direct-answer intros

First paragraph of every page answers the query in plain language. Extraction tools grab intros first.

Entity + attribute + value tables

Tables like "Model | Input price | Output price | Best for" are LLM catnip. Row-level facts are directly quotable.

Semantic HTML

One h1 per page, hierarchical h2 / h3 for sections, article > section > figure. Modern extractors respect the structure.

Consistent brand signals across sources

Same name, same sameAs URLs, same Organization schema on every domain you own. LLMs merge duplicates on strong signals.

Case study

How Cafiyn Pulse itself is optimized.

This site is the live proof of every recommendation on this page. Concrete inventory:

Signal	Where to find it	Result
FAQPage schema	325 Q&A pairs across 71 pages	Extractable answers for Google + Perplexity + ChatGPT + Claude search
HowTo schema	80 HowToSteps across 14 pages	Google how-to rich cards + step quotes in AI answers
Person (author)	Karthik Kumar on every content Article	Boosts source-credibility scoring in AI engines
Organization (publisher)	Cafiyn Innovations, cross-linked via sameAs	LLMs merge duplicates and treat as one entity
SoftwareApplication + Offer	8 tool and guide pages, price 0	Google “Free” rich card. We deliberately emit no AggregateRating: there are no real reviews to aggregate, and self-serving ratings are against Google policy, which this page lists as a mistake above
BreadcrumbList	Every deep page	Readable breadcrumbs in AI overviews
llms.txt	/llms.txt (curated TOC)	Anthropic + Perplexity source-selection hint
IndexNow	GitHub Action on every push	Bing + Yandex + DuckDuckGo re-crawl within minutes
Sitemap	/sitemap.xml, 82 URLs	Full canonical URL list for Google + Bing
Entity + attribute + value tables	Cost Comparator, this page, /free/*	Row-level quotes in AI synthesised answers
Topic cluster	/free (hub) + /free/* (pillars) + /for (hub) + /for/* (personas)	Internal PageRank concentration on the topic

Do not do this

Common AEO / GEO mistakes.

Fabricated FAQ Q&As stuffed with keywords

Google, Perplexity, and Claude all detect and demote pages where the FAQ reads like a keyword list. Write questions a real user would type; write answers a helpful human would give.

AggregateRating with fake review counts

Google can and does penalise structured-data spam. Use modest, real numbers. Only publish reviewBody with actual customer or founder statements you can defend.

Ignoring Bing because it is smaller than Google

Bing indexes ChatGPT search fallback, Copilot, and DuckDuckGo AI assist. Not being in Bing is not being in three AI engines.

One giant page instead of a topic cluster

Long-tail queries hit specific pages, not omnibus ones. Split high-value topics into a hub + pillars.

JavaScript-rendered content only

Perplexity and Claude search often skip pages that fail to render in a static fetch. Pre-render whenever the content is static.

Missing author + publisher schema

AI engines rank source credibility partly on whether they can identify the author. A named Person + Organization boost citation rate.

Engine-by-engine

What matters most on each AI answer engine.

Engine	Reads most	Wins early
Google AI Overviews	FAQPage, HowTo, Article + author, canonical entities	Strong FAQPage + fast core web vitals
Bing Copilot	Fresh Bing index (IndexNow), structured data	IndexNow-triggered recrawl + Schema.org
ChatGPT search	Bing index fallback, page freshness, entity clarity	Same as Bing + llms.txt
Perplexity	FAQPage, tables, distinctive quotable phrases, sitemap	Entity + attribute + value tables + FAQPage
Claude search	Article + author, static HTML, clean semantic structure	llms.txt + Person schema + pre-rendered pages
Gemini	Google-first (same as AI Overviews) plus structured data	Same as Google AI Overviews
DuckDuckGo AI assist	Bing index (IndexNow), structured data	IndexNow-triggered recrawl

A 30 / 60 / 90 plan

A minimal 90-day AEO / GEO plan.

The exact schedule we ran on Cafiyn Pulse, in order.

Days 1-7. Add FAQPage + HowTo + BreadcrumbList JSON-LD to every important page. Ship a real llms.txt. Wire up an IndexNow key + GitHub Action so every push re-pings Bing / Yandex / DuckDuckGo.
Days 8-30. Split every important topic into a hub + 3-5 pillar pages, each targeting a specific long-tail query. Add Article schema with author (Person) on every content page.
Days 31-60. Publish an entity + attribute + value table on your most competitive page. Add Speakable to content pages. Add sameAs cross-links between your Organization schema across every domain you own.
Days 61-90. Post the pages in relevant communities (Product Hunt, Indie Hackers, Reddit) to get first backlinks. Watch Google Search Console + Bing Webmaster Tools for the first click-throughs.

Who wrote this

About the author.

Karthik Kumar

Founder, Cafiyn Innovations. Building Cafiyn Pulse as a free front door for founders and builders. cafiyn.com

FAQ

Common questions about AEO, GEO, and SEO.

What is AEO?

AEO stands for Answer Engine Optimization. It is the practice of structuring content so answer engines (Google Featured Snippets, Bing direct answers, Amazon Alexa, Google Assistant) can extract a single correct answer to a user question and quote it, often above their traditional list of blue links. AEO leans heavily on FAQPage, HowTo, and QAPage schema, on direct-answer intro paragraphs, and on the semantic clarity of headings.

What is GEO?

GEO stands for Generative Engine Optimization. It is the practice of making your content the source that generative AI chatbots (ChatGPT search, Perplexity, Claude search, Gemini, Copilot) reach for when they synthesise an answer. GEO overlaps with AEO but adds a citation-focused layer: entity clarity, distinctive phrasing so quotes are attributable back, structured tables with entity + attribute + value, and consistent brand signals across the web.

What is the difference between SEO, AEO, and GEO?

SEO is optimizing for the ranking of a page in a list of blue links. AEO is optimizing for being the single quoted answer to a specific question. GEO is optimizing for being one of the sources a generative AI cites when it synthesises a longer answer across several sources. The three overlap heavily in 2026 and most real-world work is done together, not separately.

Do I need to choose between AEO, GEO, and SEO?

No. In 2026 the same set of underlying practices (structured data, semantic HTML, direct-answer intros, named-entity density, consistent brand signals across the web) buys you all three at once. You do not pick one; you build the fundamentals and then measure how each channel responds.

Which AI answer engines actually matter for a startup in 2026?

Ranked roughly by referral volume for indie / early-stage traffic: Google AI Overviews (largest by far), ChatGPT search (grew fastest in 2025), Perplexity, Bing Copilot, Claude search, Gemini, DuckDuckGo AI assist. Optimizing for the Bing index (via IndexNow) covers Bing Copilot, ChatGPT search fallback, and DuckDuckGo simultaneously.

What is the single highest-leverage AEO / GEO signal I can add today?

FAQPage JSON-LD with clean, direct-answer Q&A pairs. Both Google and Perplexity preferentially extract text from FAQPage.mainEntity[].acceptedAnswer.text. 5-10 real Q&As per important page will start moving needles within 2-6 weeks.

What is llms.txt and should I have one?

llms.txt is an emerging community standard for a plain-text file at your site root that gives LLM crawlers a curated table of contents of your most valuable pages, with short descriptions. It is not yet an official standard, but Anthropic, Perplexity, and several others read it. Cost to add: ~10 minutes. Ours lives at pulse.cafiyn.com/llms.txt.

What is Speakable schema and does it help?

Speakable is a SpeakableSpecification you attach to Article or WebPage schema that tells Google Assistant which parts of the page are safe to read aloud during a voice query. Currently Google-only and mostly for news, but zero cost to add and future-facing as more voice interfaces launch.

How does IndexNow help with AEO / GEO?

IndexNow is a simple protocol where you POST a list of URLs to a single endpoint and Bing, Yandex, and DuckDuckGo re-crawl them immediately. Bing’s index feeds Bing Copilot, ChatGPT search (as a fallback source), and DuckDuckGo, so IndexNow is a one-shot way to reach three answer engines. Ours runs on every push via a GitHub Action.

Does Cafiyn Pulse actually rank on these engines?

We publish this site as a live case study. Every technique in this guide is in production at pulse.cafiyn.com. Structured data, llms.txt, IndexNow, FAQPage on every important page, and 100+ direct-answer Q&A pairs across 15+ pages. Watch the results over the next 90 days and reverse-engineer what works.

Open the tool.

Every technique in this guide is deployed on this site. Use the tools and inspect the schema in the page source to reverse-engineer what worked.

See the case study live in the tools