---
title: "GEO Checklist: How to Get Cited by ChatGPT and Perplexity"
description: "A GEO checklist for startups: entity consistency, author pages, schema, citable paragraphs, earned media and a simple way to track AI citations monthly."
author: "Shilika Jain"
date: "2026-10-01T07:20:36.101+00:00"
tags: ["geo", "ai search", "seo", "pr", "checklists"]
canonical: "https://www.shilikajain.com/blog/geo-checklist-get-cited-by-chatgpt-perplexity"
---

# GEO Checklist: How to Get Cited by ChatGPT and Perplexity

By [Shilika Jain](https://www.shilikajain.com/authors/shilika-jain) - 10/1/2026

A GEO checklist for startups: entity consistency, author pages, schema, citable paragraphs, earned media and a simple way to track AI citations monthly.

---

# GEO Checklist: How to Get Cited by ChatGPT and Perplexity

To get your startup cited by ChatGPT and Perplexity, make your company easy to identify and easy to quote. Describe it the same way everywhere (site, LinkedIn, Crunchbase, press), publish pages that answer specific buyer questions in short self-contained paragraphs, add structured data and real author pages, make sure AI crawlers can reach your content, and earn mentions in publications those assistants already trust. Then measure: run the same set of buyer prompts every month and log who gets cited. Generative engine optimization (GEO) is mostly SEO discipline plus earned media, applied to a reader that's a language model.

This is the checklist version. Every item is something you can do or check this month, grouped in the order I'd tackle them.

## How AI assistants decide what to cite

It helps to know roughly what's happening under the hood. When ChatGPT, Perplexity, Claude or Google's AI Mode answer a question with web sources, they run a search (against their own index or a partner's, such as Bing), pull a handful of pages, and quote or paraphrase the passages that answer the question most directly. Sources get picked for three broad reasons:

- **Relevance.** The passage answers the exact question asked, in plain language.
- **Trust.** The domain or publication is one the system treats as credible, often established media, well-linked sites and reference sources.
- **Clarity.** The passage can be lifted cleanly: a definition, a list, a table row, a direct answer.

Separately, the underlying model has absorbed knowledge during training, largely from the same kinds of trusted sources. Consistent, widely repeated descriptions of your company make it more likely the model "knows" you at all.

Your job is to show up on all three counts. The checklist below is organised around that.

## Entity consistency: make sure the machine knows who you are

An entity is the thing an AI system recognises as your company: name, category, founders, location, what you do. If your site says "AI revenue platform," your LinkedIn says "sales intelligence," and your Crunchbase says "CRM automation," the model has three weak signals instead of one strong one.

- [ ] Write one canonical description: a single sentence on what you do and for whom. Use it everywhere.
- [ ] Use the same company name format everywhere (no mixing "Acme AI," "Acme.ai" and "AcmeAI").
- [ ] Update LinkedIn, Crunchbase, X, GitHub, Product Hunt, G2 and any directory listings to match.
- [ ] Make sure founders' bios on LinkedIn and your site name the company and role identically.
- [ ] Create or claim a Wikidata entry if your company meets their notability guidelines; don't force it if it doesn't.
- [ ] Add an About page that states founding year, founders, headquarters, category and funding plainly.
- [ ] Use the same category language in your press boilerplate as on your homepage.

If you haven't settled on one category description, do that before anything else. The [positioning statement template](/blog/positioning-statement-template-ai-startups) gets you there in an afternoon.

## Author pages and E-E-A-T signals

AI systems and Google both weigh who wrote something. Anonymous content from a young domain is easy to ignore.

- [ ] Every article has a named author with a real byline.
- [ ] Each author has a page on your site with a short bio, credentials and links to LinkedIn, X and any published work elsewhere.
- [ ] Author pages use Person structured data with sameAs links to those external profiles.
- [ ] Your founder has bylined or quoted appearances in external publications, linked from their author page.
- [ ] Technical posts are written by, or reviewed and credited to, the person who actually knows the subject.

## Structured data and technical access

If crawlers can't read your page, nothing else on this list matters.

- [ ] Content is server-rendered or static, readable without executing heavy client-side JavaScript.
- [ ] robots.txt allows the AI crawlers you want (OpenAI's, Perplexity's, Anthropic's, Google's and Bing's). Decide deliberately rather than inheriting a default block.
- [ ] Your site is submitted to both Google Search Console and Bing Webmaster Tools.
- [ ] Organization schema on the homepage with name, logo, URL, founders and sameAs profile links.
- [ ] Article schema on posts with author, datePublished and dateModified.
- [ ] Product or SoftwareApplication schema on product pages, with pricing where you publish it.
- [ ] FAQ schema only on genuine FAQ content.
- [ ] Pages load fast and don't hide core content behind tabs, accordions that require clicks, or login walls.

### What about llms.txt?

llms.txt is a proposed convention: a plain-text or markdown file at your site root that lists your most important pages with short descriptions, written for language models. Adoption by the major assistants is uneven and you shouldn't expect it alone to change citations. It's cheap to add, though, and it forces you to decide which 10 to 30 pages actually represent your company. Do it in an hour and move on.

- [ ] Publish /llms.txt listing your key pages (homepage, product, pricing, top guides, about) with one-line descriptions.
- [ ] Optionally offer clean markdown versions of key pages for agents that fetch them.

## Citable paragraphs: write for extraction

This is the highest-leverage item on the list and the one most teams skip. An assistant quotes a passage, not a page. Write passages worth quoting.

- [ ] The first paragraph of every page answers its target question directly in two to four sentences.
- [ ] Headings read like questions or search phrases ("How much does X cost" rather than "Pricing considerations").
- [ ] Each section contains at least one paragraph that makes complete sense quoted alone, with no "as mentioned above."
- [ ] Definitions follow a simple "X is a Y that does Z" pattern.
- [ ] Comparisons are in tables with clear column headers.
- [ ] Steps are in numbered or bulleted lists of five to eight items.
- [ ] Numbers are specific and sourced ("about 40% of our users," not "many users").
- [ ] Pages are dated and refreshed; stale pages lose out to recent ones.

A quick test for any paragraph: paste it alone into a message to a colleague. If they'd understand it without the rest of the page, an AI can cite it.

## Third-party mentions and earned media: the citation fuel

Your own site is one source. Assistants cross-check and often prefer independent sources, especially for "best X for Y" and "who are the leading companies in Z" questions. If the only place your company is described is your own site, you'll rarely make the list.

| Source type | Why it matters for AI citations | How to earn it |
|---|---|---|
| Tier-1 and trade press | High-trust domains frequently pulled into answers | Funding news, launches, data stories, founder commentary |
| Industry roundups and "best of" lists | Directly match comparison prompts | Pitch the authors, earn reviews, be genuinely good |
| Review sites (G2, Capterra and similar) | Structured, frequently cited for software queries | Ask happy customers systematically |
| Podcasts with published transcripts | Long, quotable founder explanations | Founder podcast tour with a clear narrative |
| Community threads (Reddit, HN, forums) | Assistants pull candid user opinions | Participate honestly; never astroturf |
| Partner and integration pages | Associates you with known brands | Co-marketing with integration partners |

This is where PR and GEO become the same job. When a founder's framing is repeated across several credible publications, it becomes the description assistants reach for. A good example is how Gaia was positioned as "the Stripe for AI agents" through a Forbes feature, a Decrypt deep-dive and a podcast tour: one consistent line, repeated by independent sources. I've written more on [why PR drives AI search citations](/blog/why-pr-drives-ai-search-citations).

- [ ] List the 10 to 20 publications your buyers and the assistants trust in your category.
- [ ] Earn at least one substantive mention per quarter from that list.
- [ ] Make sure every press mention uses your canonical description (put it in your boilerplate and press kit).
- [ ] Publish one piece of original data a year that others will cite.
- [ ] Get your founder quoted as a source on category questions, not only on company news.
- [ ] Collect reviews on the two or three review sites that rank for your category.

## Measuring AI citations

You can't manage what you don't check. Citation tracking doesn't need a fancy tool to start.

- [ ] Write 20 to 30 prompts your buyers would actually ask ("best AI tool for X," "X vs Y," "how do I solve Z").
- [ ] Run them monthly in ChatGPT (with search), Perplexity, Google AI Mode or AI Overviews, and Claude.
- [ ] Log for each prompt: are you mentioned, are you cited with a link, which competitors appear, which sources are cited.
- [ ] Track the cited sources over time. They tell you exactly which publications to pursue.
- [ ] Watch referral traffic from chatgpt.com, perplexity.ai and similar domains in your analytics.
- [ ] Ask new customers how they found you and include "AI assistant" as an option.

A simple tracking sheet:

| Prompt | ChatGPT | Perplexity | Google AI | Sources cited |
|---|---|---|---|---|
| Best [category] for [use case] | Mentioned, no link | Cited | Not mentioned | Trade pub A, review site B |
| [You] vs [competitor] | Cited | Cited | Mentioned | Your comparison page |
| How to [solve problem] | Not mentioned | Not mentioned | Not mentioned | Competitor blog, Reddit thread |

The rows where you're absent but a competitor is cited are your to-do list. Look at what got cited and earn a mention there, or publish a better answer.

## The 30-day GEO sprint

If you want a starting order, this is it.

| Week | Do |
|---|---|
| 1 | Canonical description, profile clean-up across all platforms, baseline prompt test |
| 2 | Technical access (robots, Bing, schema), author pages, llms.txt |
| 3 | Rewrite your top 10 pages for citable paragraphs and direct answers |
| 4 | Earned media plan: target publications, one data story or founder commentary pitch, review requests |

Most startups can do weeks one to three themselves. Week four is the slow, relationship-driven part, and it's the part that moves the needle most for "best X" prompts. It's also the part I run for founders, through [AI startup PR](/services/ai-startup-pr) and, for crypto teams, a [GEO and AEO programme](/pages/geo-aeo-agency-crypto) built around earned coverage. For the search fundamentals underneath all of this, see the [SEO starter guide for startups](/blog/seo-for-startups-2026-starter-guide).

You can't buy your way into an AI answer. You can make yourself the clearest, most consistently described and most independently confirmed option in your category, and that's what the assistants are looking for.

*Want the printable version with every box on one page? [Download the GEO and AI search checklist](/resources/geo-ai-search-checklist.pdf).*

---

**Book a 30-min AI visibility teardown with Shilika** - https://calendly.com/shilikajain/30min/

Canonical: https://www.shilikajain.com/blog/geo-checklist-get-cited-by-chatgpt-perplexity