Generative Engine Optimization for SaaS: How to Get Cited by ChatGPT and Perplexity

Ask ChatGPT for the best GTM agency for a SaaS startup in India and you'll get a tidy paragraph with three or four names in it. Those names got there somehow. Generative engine optimization is the work of making sure your company is one of them when a buyer asks the question that matters to you.

Below we cover what GEO and AEO mean, what the original research found, and the steps we'd take on a B2B SaaS site this month. Much GEO advice online is recycled SEO with a new acronym, so we'll flag where the evidence is thin.

What is generative engine optimization?

Generative engine optimization is the practice of making your content more likely to be retrieved, quoted and cited by AI systems that write answers. A "generative engine" is any tool that searches the web, reads several sources and writes one response with links: ChatGPT search, Perplexity, Google's AI Overviews and AI Mode, Microsoft Copilot.

The term comes from a paper titled GEO: Generative Engine Optimization, first posted to arXiv in November 2023 and published at KDD 2024, the ACM data mining conference.

What is AEO?

Answer engine optimization is the older name. It grew up around featured snippets and voice assistants, when the goal was to be the one answer Google or Alexa read out, and it puts the weight on question-and-answer formatting. Today AEO, GEO, AI search optimization and LLM SEO mean roughly the same set of tactics, and we treat AEO as the answer-formatting part of the job.

GEO vs AEO vs SEO: how do they relate?

SEO gets your page found, AEO makes an answer easy to lift from it, and GEO makes it the source an AI chooses to quote and name. They stack, and a site that skips the first layer rarely does well at the other two.

This framing is ours. The boundaries are fuzzy and the tactics overlap.

Google is blunt about the SEO dependency. Its documentation says there are "no additional requirements to appear in AI Overviews or AI Mode", and that you don't need new machine-readable files or special schema for them.

What did the original GEO research find?

The authors were Pranjal Aggarwal (IIT Delhi), Vishvak Murahari, Karthik Narasimhan and Ameet Deshpande (Princeton University), and Tanmay Rajpurohit and Ashwin Kalyan (independent researchers). They built a benchmark called GEO-bench with 10,000 queries and tested nine ways of rewriting a source page: authoritative tone, statistics addition, keyword stuffing, cite sources, quotation addition, easy-to-understand language, fluency optimization, unique words and technical terms. Visibility was scored by how much of the source appeared in the answer, weighted by position, and by a subjective impression score.

Here's what stood out for us:

• Evidence won. Quotation addition, statistics addition and cite sources were the strongest methods. The abstract reports visibility gains of up to 40%, and the best methods improved on the baseline by 41% on the word-count metric and 28% on impression.
• Keyword stuffing did worse than the untouched page. The authors write that methods "traditionally used in SEO don't perform well".
• Lower-ranked sites gained the most. Adding citations lifted visibility by 115.1% for sites ranked fifth in search results, while the top-ranked site's visibility fell by 30.3%.
• Results varied by domain. Citations helped most on factual questions, and the best pairing was fluency optimization plus statistics addition.
• It held up on Perplexity. Quotation addition improved word-count visibility by 22% there, and the methods tested showed gains of up to 37% on impression.

Keep one caveat in mind. Most experiments ran in a controlled setup built to mimic a generative engine, and real engines have changed a lot since 2023, so read the percentages as a sign of which way to lean. It matches what we see: AI answers lean on specific, attributable facts.

The third finding is good news for a young SaaS company. You can publish the page with the clearest number and the best-sourced explanation in your niche, and the research suggests that gives you a real shot at being quoted even while the category leader outranks you on Google.

How does a B2B SaaS company get mentioned in AI answers?

Steps 1 to 4 happen on your own site. Step 5 happens on other people's sites, where many mentions come from. Steps 6 to 8 are housekeeping that makes the rest count.

1. Put a direct answer at the top of every important page

AI engines skim many pages fast and pull the passage that answers the question. Open each post or landing page with two or three sentences that answer its main question completely, in plain words. This blog does exactly that: every post starts with a short answer box before the detail. Write it so it still makes sense when copied somewhere else.

2. Use question-shaped headings

Buyers type full questions into ChatGPT, and "How much does a GTM agency cost in India?" behaves very differently from the keyword "GTM agency cost". Where it reads naturally, phrase H2s as buyer questions and answer them in the first sentence below. Sales call notes, support tickets and Google's "People also ask" box will give you better questions than any keyword tool.

3. Add an FAQ section and the right schema

An FAQ block at the end of a post catches the follow-up questions that don't deserve their own heading. Mark it up with FAQPage structured data, add Article schema to each post, and put Organization schema on your homepage or about page.

Be realistic about what schema does. Google stopped showing FAQ rich results from May 2026, and it says no special schema is needed for its AI features. We still add FAQPage and Article markup to every post on this blog, because it gives machines a clear map of questions, answers, authorship and dates at no cost. Keep the marked-up text identical to the visible text.

4. Publish original numbers with named sources

This tactic has the strongest evidence behind it, and most SaaS blogs skip it. "Most SDRs take a while to ramp" gives an AI nothing to quote. "The Bridge Group's 2025 benchmark puts average SDR ramp time at 3.0 months" gives it a fact, a source and a date.

Numbers nobody else has are better still: anonymised benchmarks from your product data, a small customer survey, honest pricing breakdowns. Original data gets cited by AI engines and human writers alike, which feeds step 5. Our post on in-house vs outsourced GTM team costs is built this way, with every salary figure tied to a public source.

5. Get mentioned on sites AI engines already trust

Ask ChatGPT or Perplexity about tools in your category and look at the citations. You'll usually see the same source types: "best tools for X" listicles, G2 and Capterra, Reddit threads, LinkedIn posts and comparison articles. If your company isn't on those pages, it's unlikely to appear in the answer built from them.

This half is slow. The research and personalisation behind a good cold email for SaaS works just as well for pitching a listicle author.

6. Keep your entity information consistent

AI engines piece together who you are from many pages. If your homepage says "revenue platform", LinkedIn says "sales engagement software" and Crunchbase has a different founding year, the engine has to guess. Pick one name, one plain description and one category, and use them on your site, LinkedIn, Crunchbase, G2 and social bios. Then list those profile URLs in the sameAs field of your Organization schema.

7. Let the AI search crawlers in

An AI engine can't cite a page its crawler isn't allowed to read, and some sites blocked every AI bot when these crawlers first appeared and never revisited the decision. Check your robots.txt against each company's documentation:

We summarised these from OpenAI, Perplexity and Google crawler documentation, checked October 2026.

OpenAI's publisher FAQ tells sites to make sure they aren't blocking OAI-SearchBot, and its crawler docs say search-related robots.txt changes take about 24 hours to apply. If your CDN or firewall has bot protection, check that it isn't rejecting these agents before robots.txt even comes into play.

8. Treat llms.txt as optional

llms.txt is a proposal from Jeremy Howard, published in September 2024, for a Markdown file at /llms.txt that gives language models a short summary of your site and links to its best pages. Its status is unclear: the major AI search engines haven't publicly committed to using it, and Google says you don't need AI text files for its AI features. It takes under an hour to write, so add one if you like, after steps 1 to 7.

How do you measure AI search visibility?

Measurement is still rough, so we keep it to two simple methods.

Manual prompt tracking. Write down 20 to 30 questions your buyers actually ask, from "best [category] tool for [segment]" to "[your brand] vs [competitor]". Once a month, run each through ChatGPT and Perplexity in a fresh session and log whether you're mentioned, whether you're linked, which competitors appear and which sources get cited. Answers vary between runs, so judge the trend over a few months, and use the cited sources as your outreach list for step 5.

Referral traffic. In Google Analytics 4, filter the traffic acquisition report by session source for chatgpt.com and perplexity.ai. OpenAI says ChatGPT automatically adds utm_source=chatgpt.com to referral links. Google folds AI Overviews and AI Mode clicks into the normal "Web" search type in Search Console, so you can't split those out. Add a "how did you hear about us?" field to your demo form too, since many people read the AI answer and then type your URL directly.

Adding JSON-LD and robots.txt on a Netlify site

On a hand-written HTML site hosted on Netlify, put a plain text file called robots.txt in your publish directory (the folder Netlify deploys) so it's served at yourdomain.com/robots.txt. Here is a permissive starting point:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: *
Allow: /

Sitemap: https://www.example.com/sitemap.xml

To opt out of model training, add GPTBot and Google-Extended groups with Disallow: /. You stay in ChatGPT search and Google Search either way. After deploying, open the live URL to confirm the new file is there.

For JSON-LD, paste a <script type="application/ld+json"> block into each page's <head>. Organization schema (name, url, logo, description, sameAs) goes on the homepage. Article schema goes on each post, and FAQPage schema on any post with a visible FAQ. Check each page with the Schema Markup Validator after deploying.

What should you do first?

1. 1
   Check robots.txt and any bot protection, and unblock OAI-SearchBot and PerplexityBot.
2. 2
   Rewrite the opening of your five most important pages so each answers its main question in two or three sentences.
3. 3
   Make your company description consistent everywhere and add Organization schema.
4. 4
   Write your tracking prompts, run the first round and note which sources get cited.
5. 5
   Pitch the three most-cited third-party pages for inclusion.
6. 6
   Plan one piece of content built on numbers only you have.

Most of this is ordinary content and search work done with more care. If you're weighing whether to run it in-house or hand it to a team, our explainer on what a GTM agency does walks through the options. The GTM Fit Check also gives you a 0 to 100 score across content, outbound, paid and positioning in about a minute.

Frequently asked questions

← More from the GTM Wave blog