Free tool · Reference
A GEO checklist is the list of page, schema and access changes that answer engines reward, grouped the way the GEO score groups them. Tick off the 24 items as you fix them; progress stays in your browser, and each item links to the free helloAISearch tool that checks it.
Want this checked on every engine, on a schedule? Start the free 3-day trial. No credit card.
Nothing. Which items you tick is stored in your browser only, so the list is where you left it next time; nothing is sent to a server or to an AI provider. No AI provider is called. By using this tool you agree to the Terms and Privacy Policy.
Tick items as you fix them. Ticks are kept in this browser only; nothing is sent anywhere. Each item links to the free tool that checks it.
Engines lift the first complete answer they find. If the opening is a preamble, they quote someone else's opening.
The prompt and the H1 should match in shape. “What is X?” pages get quoted for “what is X” prompts.
Question-and-answer pairs are the easiest content for an engine to lift whole, and they match how prompts are typed.
“A vs B” prompts are answered from comparison pages. When only the competitor has written one, the engine has one side of the story.
Vague copy gets paraphrased into nothing. A price, a number and a date give the engine something to cite you for.
Engines prefer fresh sources for fast-moving categories and read dates from meta tags and JSON-LD, not from the sentence.
Two H1s leave the engine guessing what the page is about; it usually guesses the first one.
Engines lift answers section by section. A heading that states the question makes the section below it a candidate answer.
Several engines read the title and description as the page's summary before they read the body.
Without it the same content on two URLs splits the engine's confidence between them, and neither gets cited.
Some retrieval pipelines use og:title and og:description as the snippet; missing tags mean a snippet built from the first text they find.
Answer crawlers mostly do not run scripts. A page that renders client-side reads as empty to them.
A block that fails to parse is worse than none: the engine gets nothing and the validator flags the page.
This is how the engine learns what the company is. One node, reused everywhere, is what lets it merge every page into one entity.
Profiles on other sites are how the engine confirms the entity exists outside your own domain.
Schema that disagrees with the page is a trust problem; schema that matches it is a ready-made answer.
Breadcrumbs tell the engine where the page sits in the site, which is how a pricing page gets attached to the product entity.
Three spellings of the company look like three companies. Engines merge on the string.
An engine cannot cite what it is not allowed to index. A leftover noindex from staging is the most common cause of a missing page.
Blocking a search or user-fetch bot removes you from that engine's citations. Blocking a training bot does not.
Allow or block them on purpose. The mistake is a blanket rule that catches the answer crawlers too.
A short Markdown summary of the site for AI crawlers. Cheap to publish and the one file that says in your words what you are.
Crawlers that feed answers find new pages through the sitemap first. A stale lastmod tells them nothing changed.
A bot challenge, a slow origin or a redirect chain reads as a fetch failure, and a fetch failure is a page that does not exist.
Export what is left as Markdown
# GEO / SEO checklist: what is left 0 of 24 done (0%). ## Content (0/6) - [ ] The page opens with a 2–4 sentence answer to its own question, naming the brand - Why: Engines lift the first complete answer they find. If the opening is a preamble, they quote someone else's opening. - Check it: https://helloaisearch.com/tools/geo-score - [ ] The H1 is phrased the way a buyer asks, or states the answer outright - Why: The prompt and the H1 should match in shape. “What is X?” pages get quoted for “what is X” prompts. - Check it: https://helloaisearch.com/tools/geo-score - [ ] A short FAQ answers the three to six questions buyers actually ask, one paragraph each - Why: Question-and-answer pairs are the easiest content for an engine to lift whole, and they match how prompts are typed. - Check it: https://helloaisearch.com/tools/geo-score - [ ] One comparison page exists for each competitor the engines name alongside you - Why: “A vs B” prompts are answered from comparison pages. When only the competitor has written one, the engine has one side of the story. - Check it: https://helloaisearch.com/tools/compare-page-outline - [ ] The page states specifics an engine can repeat: prices, limits, dates, named features - Why: Vague copy gets paraphrased into nothing. A price, a number and a date give the engine something to cite you for. - Check it: https://helloaisearch.com/tools/ai-page-view - [ ] A visible “last updated” date, also declared in metadata - Why: Engines prefer fresh sources for fast-moving categories and read dates from meta tags and JSON-LD, not from the sentence. - Check it: https://helloaisearch.com/tools/geo-score ## Structure (0/6) - [ ] Exactly one H1 per page - Why: Two H1s leave the engine guessing what the page is about; it usually guesses the first one. - Check it: https://helloaisearch.com/tools/geo-score - [ ] Three or more H2 sections with descriptive headings, question-shaped where they answer one - Why: Engines lift answers section by section. A heading that states the question makes the section below it a candidate answer. - Check it: https://helloaisearch.com/tools/geo-score - [ ] Title tag of 30–65 characters and a meta description of 70–160 - Why: Several engines read the title and description as the page's summary before they read the body. - Check it: https://helloaisearch.com/tools/geo-score - [ ] A canonical URL on every page - Why: Without it the same content on two URLs splits the engine's confidence between them, and neither gets cited. - Check it: https://helloaisearch.com/tools/geo-score - [ ] Open Graph title and description present - Why: Some retrieval pipelines use og:title and og:description as the snippet; missing tags mean a snippet built from the first text they find. - Check it: https://helloaisearch.com/tools/geo-score - [ ] The main content is in the HTML, readable without JavaScript - Why: Answer crawlers mostly do not run scripts. A page that renders client-side reads as empty to them. - Check it: https://helloaisearch.com/tools/ai-page-view ## Schema (0/6) - [ ] JSON-LD in the head that parses without errors - Why: A block that fails to parse is worse than none: the engine gets nothing and the validator flags the page. - Check it: https://helloaisearch.com/tools/geo-score - [ ] An Organization node with name, url, logo, description and sameAs, referenced by @id from every page - Why: This is how the engine learns what the company is. One node, reused everywhere, is what lets it merge every page into one entity. - Check it: https://helloaisearch.com/tools/entity-clarity - [ ] sameAs lists at least two other profiles (LinkedIn, X, Crunchbase, GitHub, Wikidata) - Why: Profiles on other sites are how the engine confirms the entity exists outside your own domain. - Check it: https://helloaisearch.com/tools/entity-clarity - [ ] FAQPage JSON-LD that repeats the visible FAQ word for word - Why: Schema that disagrees with the page is a trust problem; schema that matches it is a ready-made answer. - Check it: https://helloaisearch.com/tools/ai-faq-schema - [ ] BreadcrumbList on inner pages - Why: Breadcrumbs tell the engine where the page sits in the site, which is how a pricing page gets attached to the product entity. - Check it: https://helloaisearch.com/tools/entity-clarity - [ ] The same brand name in the title, og:title, H1, JSON-LD and llms.txt - Why: Three spellings of the company look like three companies. Engines merge on the string. - Check it: https://helloaisearch.com/tools/entity-clarity ## Access (0/6) - [ ] No noindex on the pages you want cited - Why: An engine cannot cite what it is not allowed to index. A leftover noindex from staging is the most common cause of a missing page. - Check it: https://helloaisearch.com/tools/geo-score - [ ] robots.txt allows the crawlers that feed answers: OAI-SearchBot, ChatGPT-User, Claude-User, PerplexityBot, Googlebot - Why: Blocking a search or user-fetch bot removes you from that engine's citations. Blocking a training bot does not. - Check it: https://helloaisearch.com/tools/ai-robots-checker - [ ] A deliberate decision, written into robots.txt, on training crawlers (GPTBot, ClaudeBot, Google-Extended) - Why: Allow or block them on purpose. The mistake is a blanket rule that catches the answer crawlers too. - Check it: https://helloaisearch.com/tools/ai-robots-generator - [ ] /llms.txt published, naming the company and linking the pages that matter - Why: A short Markdown summary of the site for AI crawlers. Cheap to publish and the one file that says in your words what you are. - Check it: https://helloaisearch.com/tools/llms-txt-validator - [ ] A valid sitemap listing the canonical URLs with real lastmod dates - Why: Crawlers that feed answers find new pages through the sitemap first. A stale lastmod tells them nothing changed. - Check it: https://helloaisearch.com/tools/sitemap-validator - [ ] Pages answer 200 quickly, with no CDN rule that challenges bots - Why: A bot challenge, a slow origin or a redirect chain reads as a fetch failure, and a fetch failure is a page that does not exist. - Check it: https://helloaisearch.com/tools/geo-score Generated by the free GEO / SEO checklist at https://helloaisearch.com/tools/geo-seo-checklist.
24 items still open. The checklist tells you what to fix; tracking tells you whether it worked. helloAISearch runs your prompts on a schedule and shows which fixes moved an answer.
One check is a snapshot. A schedule is a measurement.
helloAISearch runs your prompts on ChatGPT, Claude, Gemini and Perplexity on a schedule, highlights every mention, and turns each gap into one move per day.
Questions
It is the same set of ideas as a to-do list. The GEO score fetches one page and tells you which checks it fails; the checklist is for working through a site over weeks and keeping track of what is done. Most items link to the tool that checks them.
Access first, because a blocked crawler makes the rest invisible. Then content, because an answer-first paragraph is what gets quoted. Schema and structure make what is already good easier to read. Within a group, start with the page that should win your most valuable prompt.
It removes the reasons an engine cannot cite you. Whether it does depends on the question, the competition and the day. The checklist tells you what to fix; tracking the prompts tells you whether it worked.
In your browser's local storage, on this device, under one key. Clearing site data or switching browsers resets it. The Markdown export exists so you can move the open items into your own task list.
Related tools
Read next
Free 3-day trial · No credit card · Cancel anytime
Three minutes from signup to reading what ChatGPT and Gemini say about your brand. Everything the trial finds is saved whether or not you subscribe.
hello, AI search.