SEO

AEO & GEO in 2026: How to Get Cited by ChatGPT, Perplexity & Google AI

SEO is no longer enough. Here is how to optimise your website to be found, read and cited by AI answer engines — ChatGPT, Perplexity, Gemini, Google AI Overviews. Method and concrete files for Belgian SMEs.

Published
Updated
10 min
1,919words
#AEO#GEO#ChatGPT SEO#Perplexity#Google AI Overviews#answer engine optimization#generative engine optimization#llms.txt#GPTBot

Why classic SEO is no longer enough

For 25 years, SEO had one simple goal: rank a web page in Google's 10 blue links. In 2026, the equation has shifted. A growing share of searches now resolve without a click: the user asks ChatGPT, Perplexity, Gemini, or sees the answer directly in Google AI Overviews, and the page is cited as a source — sometimes not at all.

Two new disciplines have emerged:

  • AEO (Answer Engine Optimization) — optimise to be the answer returned by an answer engine.
  • GEO (Generative Engine Optimization) — optimise to be cited as a source in a generated LLM answer. The term was formalised in November 2023 in the academic paper "GEO: Generative Engine Optimization" (Aggarwal et al., arXiv:2311.09735), now widely used in the industry.
AEO/GEO do not replace SEO. The technical and editorial foundations of SEO (speed, structure, depth, authority) remain essential. AEO/GEO add a layer: making your content legible, trustworthy and citable by a machine that synthesises.

How ChatGPT, Perplexity and Google AI find information

To optimise, you must first understand how each engine retrieves its sources:
EngineInformation sourceCrawler to allow
ChatGPT (search / browsing)Own web index + Bing as historical search partnerOAI-SearchBot, ChatGPT-User, GPTBot
ChatGPT (model training)OpenAI crawlGPTBot
PerplexityOwn crawl + web indexPerplexityBot, Perplexity-User
Google AI Overviews / Gemini (Search)Google indexGooglebot (allow) + Google-Extended (separate AI consent)
Claude (Anthropic)Own crawl + trainingClaudeBot, Claude-User, anthropic-ai
Bing Chat / CopilotBing indexBingbot
Common Crawl (used to train many LLMs)Public crawlCCBot
The first concrete AEO/GEO action: know which bots you allow in your robots.txt. Blocking GPTBot removes your content from future OpenAI training; blocking OAI-SearchBot or ChatGPT-User removes it from ChatGPT's answers (two different things).

A 2026-ready robots.txt

Here is a robots.txt that allows all major AI engines for reading and explicitly states it — recommended for an SME that wants maximum visibility:
# Search & AI Search engines
User-agent: Googlebot
Allow: /
User-agent: Bingbot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Perplexity-User
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: Claude-User
Allow: /

# Training crawlers — allow or block depending on your strategy
User-agent: GPTBot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: anthropic-ai
Allow: /
User-agent: CCBot
Allow: /

Sitemap: https://www.example.be/sitemap.xml
If you prefer to be cited in answers without your content feeding training, allow the search bots (OAI-SearchBot, Perplexity-User, ChatGPT-User) and block the training bots (GPTBot, Google-Extended, anthropic-ai).

llms.txt: the new convention to describe your site to AIs

In September 2024, Jeremy Howard (co-founder of fast.ai and answer.ai) proposed the /llms.txt standard — a Markdown file at the site root that summarises the content and provides links to key pages in a form directly digestible by an LLM. Spec: llmstxt.org.

The principle: an LLM has a limited context window. Giving it a concise Markdown map of your site improves the quality of answers that cite your content.

Minimal example for a Brussels SME:

# LeadCo — Web Agency in Brussels

> LeadCo is a Belgian digital agency based in Brussels, specialised
> in website creation and SEO for SMEs and freelancers.

## Services
- [Website creation](https://www.leadco.be/en/services/creation-site)
- [SEO](https://www.leadco.be/en/services/seo)
- [E-commerce](https://www.leadco.be/en/services/ecommerce)

## Reference articles
- [How much does a website cost in Belgium](https://www.leadco.be/en/blog/prix-site-internet-belgique)
- [Local SEO in Brussels](https://www.leadco.be/en/blog/seo-local-bruxelles)
- [GDPR and websites in Belgium](https://www.leadco.be/en/blog/rgpd-site-web-belgique)

## Contact
- Email: hello@leadco.be
- Brussels, Belgium
This is an emerging convention, not an official W3C/IETF standard. But it is already adopted by Anthropic (docs.anthropic.com/llms.txt), Vercel, Cloudflare, Cursor and many agencies. Implementation cost is essentially zero; potential upside (better LLM understanding) is real.

Content patterns that get cited

The "GEO: Generative Engine Optimization" paper (Aggarwal et al., 2023) compared, on 10,000 queries, different methods to increase visibility in generative answers. Patterns that statistically work best:
  1. In-text source citations ("according to a study by X…")
  2. Precise numerical statistics
  3. Direct quotes from experts (attributed quotes)
  4. Clear, authoritative language (no keyword stuffing)
Conversely, techniques that did not improve, or actively decreased, visibility:
  • Keyword stuffing
  • "Too fluid" prose without data
  • Unique text without structure

Concrete editorial implications

  • A short definition at the top of every article. An AI looks for the sentence that answers "what is X".
  • A self-contained answer per section. Each H2/H3 must be readable out of context and provide a complete answer.
  • Numbers and dates. An AI is more likely to cite a sentence like "in July 2023, the EU-US Data Privacy Framework entered into force" than a vague paragraph on compliance.
  • One question, one answer. FAQ sections with H3 = question, paragraph = answer, are among the most cited content by Perplexity and AI Overviews.
  • Verifiable external sources. Links to official publications, numbered studies, primary documentation. An LLM weights more heavily a text that cites its sources.

Structured data: indispensable for AEO

Schema.org remains the common language between Google, Bing and the LLMs. For a Belgian SME, the most useful types:
SchemaWhen to use
Organization"About" page — your identity, logo, social profiles
LocalBusiness (or sub-types: ProfessionalService, Restaurant…)Any business with a local address
WebSite + SearchActionEnable Google sitelinks search box
BreadcrumbListImproves SERP display
Article / BlogPostingAny editorial article
FAQPageFAQ sections — eligible for rich results
HowToStep-by-step guides
Product + Offer + AggregateRatingE-commerce
ServiceService pages
PersonAuthor pages (useful for E-E-A-T)
Structured data is read by AI crawlers as well as Google. It disambiguates what your page is saying: "this is an article, written by X, on date Y, about topic Z."

Free verification tools:

E-E-A-T: what Google reinforced in December 2022

Google's Search Quality Rater Guidelines added an extra E in December 2022: Experience, alongside Expertise, Authoritativeness, Trustworthiness. LLMs, which lean on Google and Bing indexes, inherit the same logic.

To demonstrate E-E-A-T on an SME site:

  • Identified author on each article (dedicated author page, photo, bio)
  • Person schema linking to LinkedIn / publications
  • Publication AND update date explicit
  • Sources cited systematically in articles
  • External mentions (press, podcasts, talks) listed on the "About" page
  • Clear legal identity (legal notice, company number, verifiable address)

What works concretely, per engine

Google AI Overviews / Gemini

  • Standard Google optimisation (Core Web Vitals, schema, quality content)
  • Cited sites are usually those already in the organic top 5
  • Tip: structure as key paragraph then details — Google often extracts the first paragraph under an H2/H3 as the reference passage
  • Presence in Bing's index helps (use Bing Webmaster Tools)
  • OAI-SearchBot not blocked in robots.txt
  • Pages with direct, structured answers — lists, tables, definitions

Perplexity

  • Own indexing + Google + Bing
  • Favours fresh, numerically rich sources; a recent article with explicit dates is more likely to be reused
  • PerplexityBot and Perplexity-User allowed

ChatGPT (model, no browsing)

  • Much harder: you must be in training sources (Common Crawl, Wikipedia dumps, etc.) or sufficiently cited for the model to have learned your brand
  • This is where digital PR (Wikipedia mentions, press, reference blogs, podcasts) regains its full value

IndexNow: the quiet bonus for Bing & Yandex

IndexNow (indexnow.org) is an open protocol co-developed by Microsoft (Bing) and Yandex in 2021. It allows a site to instantly notify participating search engines of a new or updated URL, instead of waiting for a re-crawl.

For content sites (active blog, e-commerce with stock changes), IndexNow significantly accelerates Bing-side indexing — and therefore ChatGPT search visibility. Most modern CMS and frameworks (Next.js with a plugin, WordPress with Rank Math or Yoast, Shopify, Wix) support IndexNow in a few clicks.

Measuring AI engine visibility

Tracking is harder than classic SEO, but several signals are available:

  • Traffic from Perplexity, ChatGPT, Gemini: visible in Google Analytics 4 under sources perplexity.ai, chat.openai.com, gemini.google.com. Create a custom channel to track these referrals.
  • Brand mentions: regularly search your brand in ChatGPT, Perplexity, Gemini. Note the cited sources. Dedicated tools: Otterly.ai, Profound, AthenaHQ (recent, with limited free plans).
  • Search Console: track AI Overviews appearance via CTR-vs-impression evolution on informational queries.

AEO/GEO action plan in 8 steps

For a Belgian SME in 2026, the order that maximises ROI:

  1. Audit robots.txt — allow desired AI bots, block undesired ones, declare the sitemap
  2. Publish /llms.txt — short, up to date, listing key pages
  3. Implement priority schemas: Organization, LocalBusiness, Article, FAQPage, BreadcrumbList
  4. Restructure each content page: short definition first, self-contained sections, FAQ at the bottom
  5. Strengthen E-E-A-T: author pages, update dates, cited sources
  6. Enable IndexNow to notify Bing & Yandex
  7. Submit the site to Bing Webmaster Tools (and IndexNow at the same time)
  8. Set up tracking: GA4 channel for AI referrers + monthly citation review
None of these 8 actions costs more than a day of implementation on a standard site. The payoff: being cited by default when a prospect asks ChatGPT "which web agency in Brussels would you recommend?"

FAQ

Are AEO and GEO the same thing?

Not exactly. AEO (Answer Engine Optimization) is broader and covers any answer engine — including Google featured snippets, Alexa, Siri. GEO (Generative Engine Optimization) is specific to generative LLM-based engines (ChatGPT, Perplexity, Gemini, Claude). In 2026, the two converge in practice.

If I block GPTBot, do I disappear from ChatGPT?

No. GPTBot is used for training OpenAI models. For search inside ChatGPT (the "browsing" feature), OAI-SearchBot and ChatGPT-User are involved. You can block one and allow the other.

Is the llms.txt file mandatory?

No, it's an emerging convention. But it's trivial to set up and several major publishers (Anthropic, Vercel, Cloudflare) use it. Near-zero cost, real potential upside: better LLM understanding of your content.

How long before AEO/GEO results?

Faster than classic SEO on recent AI engines (Perplexity, ChatGPT search): citations appear within 2 to 6 weeks for niche topics. On Google AI Overviews, the timeframe aligns with classic SEO (3 to 6 months).

Does traditional SEO become useless?

On the contrary: a site that does not rank in SEO does not rank in AEO either. Generative engines lean on existing indexes (Google, Bing) and their own crawlers, which broadly apply the same criteria: authority, structure, editorial quality.

Are there free tools to track AI visibility?

A few angles: monitor perplexity.ai, chat.openai.com, gemini.google.com referrers in Google Analytics 4; manually search your brand in each AI engine monthly and note cited sources. For automated tracking, several dedicated tools (Otterly.ai, Profound, AthenaHQ) offer limited free plans.
The takeaway: AEO and GEO do not replace SEO, they extend it. The foundations remain the same — speed, structure, useful content, authority — but the upper layer changes: you must now write for summarising machines as much as for scanning humans. Belgian SMEs that start in 2026 will gain a durable lead over competitors who wait.
LeadCo

LeadCo Agency

Founder of LeadCo

Web developer and conversion expert based in Brussels. I build websites that turn your visitors into paying customers.

Discover our services

Need a website that converts?

Let’s discuss your project for free

Request a quote