# aivisibilityskill: full reference > AI visibility for coding agents. An API, MCP server and agent skill that let Claude Code, Codex or Cursor see what ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok and DeepSeek say about a product, get a fix plan to run in the product's own codebase, and re-check whether the fix worked. - Status: open to everyone. Sign up at https://app.aivisibilityskill.com/login, or run a first check in the browser at https://app.aivisibilityskill.com/start - Pricing: prepaid credits, $1 = 1,000 credits on packs; 500 free credits at signup (30 days); plans $9 (10,000 credits), $29 (35,000), $99 (130,000), $299 (400,000) per month; yearly plans 15% off - An answer costs 11 to 25 credits on ChatGPT, Gemini, Google AI Overviews and Google AI Mode (batch, standard or live); DeepSeek 25, Perplexity 35, Claude 60, Grok 200 - Every answer keeps the raw engine response as a verifiable receipt (SHA-256). Failed work is not charged - Made by Ahmed Gagan (https://ahmedgagan.in, https://x.com/ahmedgagan11), maker of DataBull Last updated: 2026-11-03. Index of this file's sources: https://aivisibilityskill.com/llms.txt ## How it works 1. See it: ask ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok and DeepSeek the questions buyers ask; get every answer, the brands each one names and every source it cites, and how often the product is named against competitors. 2. Fix it: a site check reads the product's site the way AI crawlers do (robots.txt rules per crawler, pages empty without JavaScript, structured data, llms.txt, prices); research reads the pages engines cite; problems are ranked with evidence; a fix plan tells the coding agent exactly what to change in the codebase. 3. Prove it: a live check confirms the fix is on the site, and the engines are asked again later for a before-and-after. ## Ways to use it - Browser: https://app.aivisibilityskill.com/start suggests buyer questions and competitors from the homepage, runs 3 questions on 4 engines, a site check and a problem scan (about 350 credits, inside the 500 free signup credits). - MCP server: https://mcp.aivisibilityskill.com/mcp (OAuth). Claude Code: `claude mcp add --transport http aivisibility https://mcp.aivisibilityskill.com/mcp`. Codex: `codex mcp add aivisibility --url https://mcp.aivisibilityskill.com/mcp && codex mcp login aivisibility`. Cursor, VS Code, ChatGPT: add the URL as a custom MCP server. - REST API: https://api.aivisibilityskill.com, Bearer keys (aiv_live_…, or aiv_test_… for free sample answers). POST /v1/checks, GET /v1/jobs/{id}, POST /v1/estimate (free), projects, visibility, competitors, sources, site checks, opportunities, fix plans and verification. Reference: https://api.aivisibilityskill.com/docs ## Credits | What | Credits | |---|---| | One answer from ChatGPT, Gemini or Google AI (batch / standard / live) | 11 / 15 / 25 | | One answer from DeepSeek / Perplexity / Claude / Grok | 25 / 35 / 60 / 200 | | Site check (reading the site the way AI crawlers do) | 20 | | Reading the pages engines cite for a question | 40 to 70 | | Ranked problems for a product | 30 | | A fix plan | 40 | | Checking a fix is live | 20 | | Reading results, estimates | free | Plan credits last 60 days, packs 12 months, signup credits 30 days. Agents have a daily limit and ask before spending more than 300 credits at once. ## Comparisons ## Profound alternative Profound is an enterprise AI visibility platform: after a 7-day free trial, the paid plan is Enterprise, priced through sales, and that's where its API, exports and MCP server are. If you're a founder or small team who wants to start today, aivisibilityskill is self-serve, starts with 500 free credits, charges per answer, and lets your coding agent fix what it finds. - Best for: Profound: Enterprise marketing teams and agencies. aivisibilityskill: Founders, developers and small agencies who work with a coding agent. - How you use it: Profound: Web platform, with an MCP server on Enterprise. aivisibilityskill: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console. - Price to start: Profound: 7-day free trial: 50 prompts a day on 3 engines, no API, exports or history. aivisibilityskill: Free: 500 credits, no card. - Paid plans: Profound: Enterprise, price on request. aivisibilityskill: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok. - AI engines: Profound: Up to 9 on Enterprise, including ChatGPT, Perplexity, Gemini, Google AI Overviews, AI Mode, Copilot, Claude, DeepSeek and Exa. aivisibilityskill: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek. - API: Profound: Enterprise plan. aivisibilityskill: Every plan, including free credits. - MCP server for agents: Profound: Enterprise plan. aivisibilityskill: Every plan, sign in with OAuth. - Fixes: Profound: Suggestions and content tools inside the platform. aivisibilityskill: Ranked problems, then a fix plan your agent applies in your codebase and re-checks. - Proof of each answer: Profound: Not listed on its pricing page. aivisibilityskill: The raw engine response is kept with a SHA-256 hash. Sources: https://www.tryprofound.com/pricing, https://docs.tryprofound.com/mcp/overview ## Peec AI alternative Peec AI is a tracking dashboard for SEO and marketing teams, from $95 a month for 50 prompts on 3 active AI models, with an MCP server on every plan and the API from $245 a month. aivisibilityskill starts free, charges per answer across 8 engines, includes the API on every plan, and lets your coding agent fix what it finds. - Best for: Peec AI: SEO and marketing teams tracking a set of prompts every day. aivisibilityskill: Founders, developers and small agencies who work with a coding agent. - How you use it: Peec AI: Web dashboard, plus an MCP server. aivisibilityskill: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console. - Price to start: Peec AI: Starter at $95 a month: 50 prompts, 1 country, 3 active models. aivisibilityskill: Free: 500 credits, no card. - Paid plans: Peec AI: $95, $245 or $495 a month; Enterprise on request. aivisibilityskill: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok. - AI engines: Peec AI: 3 active models at once on paid plans (ChatGPT, Gemini, Google AI Mode, AI Overviews, Copilot and more); unlimited on Enterprise. aivisibilityskill: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek. - API: Peec AI: Pro ($245 a month) and up. aivisibilityskill: Every plan, including free credits. - MCP server for agents: Peec AI: Every plan. aivisibilityskill: Every plan, sign in with OAuth. - Fixes: Peec AI: Recommendations for your team to act on. aivisibilityskill: Ranked problems, then a fix plan your agent applies in your codebase and re-checks. - Proof of each answer: Peec AI: Not listed on its pricing page. aivisibilityskill: The raw engine response is kept with a SHA-256 hash. Sources: https://peec.ai/pricing, https://docs.peec.ai/mcp/introduction ## Scrunch alternative Scrunch is an AI visibility platform for brand and agency teams. Its Core plan is $250 a month for 125 prompts on 4 engines, and its API, MCP access and Agent Experience Platform are on Enterprise. aivisibilityskill starts free, covers 8 engines on every plan, includes the API and MCP server, and fixes your site through your coding agent. - Best for: Scrunch: Brand and agency marketing teams. aivisibilityskill: Founders, developers and small agencies who work with a coding agent. - How you use it: Scrunch: Web dashboard; Agent Experience Platform on Enterprise. aivisibilityskill: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console. - Price to start: Scrunch: 7-day free trial, then Core at $250 a month (125 prompts, 5,000 responses). aivisibilityskill: Free: 500 credits, no card. - Paid plans: Scrunch: Core $250 a month; Enterprise on request. aivisibilityskill: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok. - AI engines: Scrunch: 4 on Core (ChatGPT, Perplexity, Google AI Overviews, Copilot); 9 on Enterprise, adding Claude, Gemini, Meta AI, AI Mode and Grok. aivisibilityskill: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek. - API: Scrunch: Enterprise plan. aivisibilityskill: Every plan, including free credits. - MCP server for agents: Scrunch: Enterprise plan. aivisibilityskill: Every plan, sign in with OAuth. - Fixes: Scrunch: Site audits, page optimizations and content generation. aivisibilityskill: Ranked problems, then a fix plan your agent applies in your codebase and re-checks. - Proof of each answer: Scrunch: Not listed on its pricing page. aivisibilityskill: The raw engine response is kept with a SHA-256 hash. Sources: https://scrunch.com/pricing, https://scrunch.com/platform/agent-experience/ ## AthenaHQ alternative AthenaHQ is a GEO platform for marketing teams with a free plan of 300 credits and a Starter plan at $295 a month for 3,600 credits across 11 engines; API access is a paid add-on. aivisibilityskill fills the gap between the two with plans from $9 a month, includes the API and an MCP server on every plan, and fixes your site through your coding agent. - Best for: AthenaHQ: Marketing teams at brands and enterprises. aivisibilityskill: Founders, developers and small agencies who work with a coding agent. - How you use it: AthenaHQ: Web platform with its own AI agent. aivisibilityskill: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console. - Price to start: AthenaHQ: Free Essential plan: 300 credits on 5 engines. aivisibilityskill: Free: 500 credits, no card. - Paid plans: AthenaHQ: Starter $295 a month for 3,600 credits; Enterprise on request. aivisibilityskill: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok. - AI engines: AthenaHQ: 11 on Starter, including ChatGPT, Perplexity, Google AI Overviews, AI Mode, Gemini, Claude, Copilot, Grok, DeepSeek, Meta AI and Mistral. aivisibilityskill: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek. - API: AthenaHQ: Paid add-on on Starter; included on Enterprise. aivisibilityskill: Every plan, including free credits. - MCP server for agents: AthenaHQ: None listed on its site. aivisibilityskill: Every plan, sign in with OAuth. - Fixes: AthenaHQ: Recommendations and content agents inside the platform. aivisibilityskill: Ranked problems, then a fix plan your agent applies in your codebase and re-checks. - Proof of each answer: AthenaHQ: Not listed on its pricing page. aivisibilityskill: The raw engine response is kept with a SHA-256 hash. Sources: https://www.athenahq.ai/pricing ## Otterly.AI alternative Otterly.AI is an AI search monitoring tool from $29 a month for 15 prompts. ChatGPT, Google AI Overviews, Perplexity and Copilot are included; Gemini, AI Mode and Claude are paid add-ons, and the API and MCP start on the $189 Standard plan. aivisibilityskill charges per answer across 8 engines on every plan and lets your coding agent fix what it finds. - Best for: Otterly.AI: Marketers and agencies tracking prompts daily. aivisibilityskill: Founders, developers and small agencies who work with a coding agent. - How you use it: Otterly.AI: Web dashboard and Semrush app; API and MCP on Standard and up. aivisibilityskill: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console. - Price to start: Otterly.AI: Lite at $29 a month for 15 prompts, with a free trial. aivisibilityskill: Free: 500 credits, no card. - Paid plans: Otterly.AI: $29, $189 or $489 a month (less on annual billing); Enterprise on request. aivisibilityskill: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok. - AI engines: Otterly.AI: ChatGPT, Google AI Overviews, Perplexity and Copilot included; Gemini, AI Mode and Claude are paid add-ons. aivisibilityskill: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek. - API: Otterly.AI: Standard and up (2,000 requests a month). aivisibilityskill: Every plan, including free credits. - MCP server for agents: Otterly.AI: Standard and up (2,000 requests a month). aivisibilityskill: Every plan, sign in with OAuth. - Fixes: Otterly.AI: GEO audits and recommendations. aivisibilityskill: Ranked problems, then a fix plan your agent applies in your codebase and re-checks. - Proof of each answer: Otterly.AI: Not listed on its pricing page. aivisibilityskill: The raw engine response is kept with a SHA-256 hash. Sources: https://otterly.ai/pricing aivisibilityskill: bestFor: Founders, developers and small agencies who work with a coding agent; workflow: Inside Claude Code, Cursor, Codex or ChatGPT, plus a light web console; start: Free: 500 credits, no card; pricing: $9, $29, $99 or $299 a month, or $10 credit packs. An answer costs 11 to 25 credits (about 1 to 2.5 cents) on ChatGPT, Gemini and Google; more on Claude and Grok; engines: 8 on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok, DeepSeek; api: Every plan, including free credits; mcp: Every plan, sign in with OAuth; fixes: Ranked problems, then a fix plan your agent applies in your codebase and re-checks; proof: The raw engine response is kept with a SHA-256 hash. Competitor details checked 2026-11-03. ## What AI engines cite when buyers ask what to buy Across 2,381 answers to 300 software buyer questions, four AI engines cited 3,960 different websites. YouTube, Reddit and Zapier were cited most, two thirds of the cited pages were “best of”, “vs” or review pages, and 71% of the sites cited for a question were cited by only one engine. ### Key findings - Google's AI leans on YouTube and Reddit. ChatGPT doesn't. Google AI Overviews cited YouTube in 85% of its answers and Reddit in 48%. ChatGPT cited YouTube once and Reddit 3 times in 5,986 citations. - The engines rarely agree. On average, 71% of the sites cited for a question were cited by only one of the four engines. Only 106 of 298 questions had even one site that all four cited. - Ask twice, get different sources. Asking the same engine the same question twice, the two answers shared 46% of their cited sites on average. ChatGPT never cited exactly the same set of sites twice (0 of 298 questions). - Engines cite roundups. 67% of cited pages had “best”, “top”, “vs”, “alternatives”, “compare” or “review” in the title. For Gemini it was 95%. According to aivisibilityskill's analysis of 24,258 citations, 71% of the websites AI engines cite for a buyer question are cited by only one engine. Checking one engine, once, shows you a small and unstable slice of what buyers see. aivisibilityskill, What AI engines cite, 300 questions × 4 engines × 2 samples, October 29 to November 1, 2026 ### How we collected the data We asked 300 buyer questions about software, such as “best accounting software for small business”, “1password vs bitwarden” and “ahrefs alternatives”, to ChatGPT (with web search), Gemini, Google AI Overviews and Google AI Mode. Each engine got each question twice, in US English, without any account history. Answers were collected between October 29 and November 1, 2026, through the same pipeline the product uses. A citation is a link the answer shows as a source. We counted websites by domain, with “www.” removed. Gemini answered 587 of its 600 prompts and Google showed an AI Overview for 594 of 600; the rest are excluded. Limits: software categories only, one country and language, a four-day window, and two samples per question. Treat the numbers as a snapshot of how these engines behave for software buyers, not a law. ### Each engine, side by side Citations by engine, 300 buyer questions asked twice, Oct 29 to Nov 1, 2026 EngineAnswersCitationsPer answerRoundup titlesYouTubeReddit ChatGPT6005,98610.041%1 of 5,986 citations3 of 5,986 citations Gemini5878,62814.795%63 of 8,628 citations84 of 8,628 citations Google AI Overviews5944,9078.364%85% of answers48% of answers Google AI Mode6004,7377.952%50% of answers45% of answers Gemini cites the most sources per answer, and nearly all of them are roundups. Google's two AI surfaces behave alike, and differently from ChatGPT: video and forum threads for Google, written roundups for ChatGPT. ### The sites cited most Share of all 2,381 answers that cite each site at least once SiteAnswers citing itShare of answers youtube.com84735.6% reddit.com60725.5% zapier.com59224.9% pcmag.com2038.5% techradar.com1436.0% itechguides.com1144.8% nerdwallet.com1114.7% forbes.com1104.6% gartner.com994.2% security.org954.0% Two things stand out. First, general-purpose sites (YouTube, Reddit) dominate because Google's AI uses them so heavily. Second, a handful of roundup publishers carry a lot of weight: itechguides.com appeared in 114 answers, all of them ChatGPT's. If a roundup like that leaves you out, the engine that relies on it probably will too. ### What to do about it - Check more than one engine, more than once. With so little overlap, a single ChatGPT chat tells you little. Measure how often you're named across engines and samples. Here's how. - Find the roundups that decide your category. Look at which pages the engines cite for your buyers' questions, then get onto those lists or publish an honest comparison page of your own. - For Google's AI, think beyond your website. A clear YouTube walkthrough and real answers in the Reddit threads Google cites can matter as much as your own pages. - Make sure the crawlers can read you. If OAI-SearchBot or PerplexityBot is blocked, the engines can't cite you however good the page is. See which AI crawlers to allow. ### Cite this study You're welcome to use these numbers. Please link to this page: aivisibilityskill (2026). What AI engines cite. https://aivisibilityskill.com/guides/what-ai-engines-cite Q: Which websites does ChatGPT cite most? A: In our 300 software buyer questions, ChatGPT cited zapier.com, itechguides.com, rankite.com, dupple.com and techradar.com most often. It almost never cited YouTube or Reddit: once and 3 times in 5,986 citations. Q: Does Google AI Overviews cite YouTube and Reddit? A: Heavily. Google AI Overviews cited YouTube in 85% of its answers and Reddit in 48%. Google AI Mode cited YouTube in 50% and Reddit in 45%. Q: Do ChatGPT, Gemini and Google's AI cite the same sources? A: Rarely. On average 71% of the sites cited for a question were cited by only one of the four engines, and only 106 of 298 questions had a site that all four cited. Q: Why do AI answers change when I ask again? A: AI engines search the web each time and sample their answers, so sources vary. Asking the same engine the same question twice, the answers shared 46% of their cited sites on average. ## Which AI crawlers to allow in robots.txt To show up in AI answers, allow the search crawlers: OAI-SearchBot for ChatGPT search, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot. Blocking training crawlers such as GPTBot, ClaudeBot or CCBot keeps your pages out of model training, but doesn't remove you from AI search answers. ### Three kinds of AI crawler - Search crawlers build the index an AI engine searches before it answers. Block them and the engine can't cite you. Examples: OAI-SearchBot, PerplexityBot, Claude-SearchBot, and Google's and Bing's normal crawlers. - User fetchers open a page because a person asked the assistant to look at it. Examples: ChatGPT-User, Perplexity-User, Claude-User. - Training crawlers collect pages to train future models. Examples: GPTBot, ClaudeBot, CCBot, and the Google-Extended control. Blocking them is a legitimate choice, and it doesn't take you out of AI search. ### Every crawler, and what blocking it does AI and search crawlers, as checked by the aivisibilityskill site check User agentOperatorKindIf you block it GooglebotGoogleSearchGoogle Search, including AI Overviews and AI Mode, can't crawl these pages. BingbotMicrosoftSearchBing can't crawl these pages; Bing's index feeds Copilot and other AI search tools. OAI-SearchBotOpenAISearchChatGPT search can't show or link these pages in its answers. ChatGPT-UserOpenAIUser fetchChatGPT can't open these pages when a user asks it to look something up. PerplexityBotPerplexitySearchPerplexity can't index these pages for its answers. Perplexity-UserPerplexityUser fetchPerplexity can't open these pages while answering a user. Claude-SearchBotAnthropicSearchClaude's web search can't index these pages. Claude-UserAnthropicUser fetchClaude can't open these pages while answering a user. Google-ExtendedGoogleTrainingGemini apps may not use these pages for grounding or training. Google Search and AI Overviews are not affected. GPTBotOpenAITrainingOpenAI won't use these pages to train models. ChatGPT search is controlled by OAI-SearchBot, not this. ClaudeBotAnthropicTrainingAnthropic won't use these pages to train models. CCBotCommon CrawlTrainingCommon Crawl (used to train many models) won't archive these pages. According to aivisibilityskill, the most common robots.txt mistake we see is blocking GPTBot and assuming that's all it does. ChatGPT search is controlled by a different crawler, OAI-SearchBot, and blocking that one is what keeps you out of ChatGPT's answers. From the aivisibilityskill site check, which reads robots.txt the way each crawler does ### robots.txt you can copy Allow everything (the default when there's no rule, and the right choice for most products that want to be recommended): User-agent: * Allow: / Sitemap: https://yoursite.com/sitemap.xml Stay in AI search, opt out of training: User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: CCBot Disallow: / User-agent: Google-Extended Disallow: / User-agent: * Allow: / Sitemap: https://yoursite.com/sitemap.xml Keep private areas out for everyone with a path rule (for example Disallow: /app/) rather than blocking whole crawlers. ### Mistakes that hide you from AI answers - Blocking OAI-SearchBot or PerplexityBot by accident, often by copying a “block all AI bots” list that mixes search and training crawlers. - A firewall or CDN rule that blocks AI bots even though robots.txt allows them. Check your bot protection settings too. - Pages that are empty without JavaScript. Many crawlers don't run JavaScript, so a client-rendered page can look blank to them. - A robots.txt that errors. If /robots.txt returns a server error, crawlers may treat the whole site as off limits for a while. ### Check your site The free check includes a site check that reads your robots.txt the way each crawler above does, finds pages that are empty without JavaScript, and tells your coding agent exactly what to change. You can also run it from your agent with the MCP server; it costs 20 credits. Q: Does blocking GPTBot remove my site from ChatGPT? A: No. GPTBot collects pages for training. ChatGPT search uses OAI-SearchBot, and ChatGPT-User opens pages when a person asks. Block GPTBot if you don't want your pages used for training; leave OAI-SearchBot allowed if you want ChatGPT to cite you. Q: What is OAI-SearchBot? A: OAI-SearchBot is OpenAI's search crawler. It builds the index ChatGPT search uses to find and link pages. If robots.txt blocks it, ChatGPT search can't show or link your pages in its answers. Q: Does Google-Extended affect AI Overviews? A: No. Google-Extended controls whether Gemini apps may use your pages for grounding or training. Google Search, including AI Overviews and AI Mode, uses Googlebot. Q: Should I block AI crawlers? A: Block training crawlers if you don't want your content used to train models. Don't block search crawlers if you want AI answers to recommend you: they're how engines find and cite your pages. ## How to check if ChatGPT recommends your product Write down the questions your buyers ask, ask each engine every question more than once without your own history, and count how often you're named and which sites are cited. One ChatGPT chat is a sample of one: in our data, asking the same question twice gave answers that shared only 46% of their sources. ### Why one chat isn't enough ChatGPT and other engines search the web each time and sample their answers. In our study of 24,258 citations, ChatGPT never cited exactly the same set of sites twice for the same question, and 71% of the sites cited for a question were cited by only one engine. A single chat in your own logged-in account, shaped by your history, is the least reliable sample you can take. ### Step by step - List your buyers' questionsWrite 5 to 10 questions the way buyers type them: “best invoicing app for freelancers”, “freshbooks alternatives”, “invoice software for small agencies”. Leave your brand name out. - Ask more than one engineChatGPT, Gemini, Google AI Overviews and AI Mode cite different sites. Check at least those four. - Ask each question more than onceAnswers vary between runs. Ask at least twice, logged out or in a private window, so your own history doesn't shape the answer. - Record three thingsWhether you're named, which competitors are named instead, and which sites the answer cites as sources. - Read the cited pagesThe roundups and comparisons the engines cite usually decide who gets recommended. Note which ones leave you out. - Fix the biggest gap, then check againGet onto a cited roundup, publish a comparison page, fix a wrong price or unblock a crawler. Re-run the same questions a week or two later. ### What to record An example sheet: one row per answer QuestionEngineRunYou named?Named insteadCited sites best invoicing app for freelancersChatGPT1NoFreshBooks, Wave, Zoho Invoicenerdwallet.com, wise.com best invoicing app for freelancersChatGPT2NoFreshBooks, Wave, Bonsaihostinger.com, nerdwallet.com best invoicing app for freelancersGoogle AI Overviews1NoFreshBooks, Zoho Invoiceyoutube.com, reddit.com Your mention rate is the share of rows where you're named. Track it per engine; one engine can name you while another never does. ### Doing it automatically Doing this by hand for 10 questions, 4 engines and 2 runs is 80 chats. aivisibilityskill runs the same check for you: - In your browser: type your site at app.aivisibilityskill.com/start. It suggests buyer questions and competitors from your homepage, asks the engines, and shows who's named, which sites are cited and what to fix first. The first check fits in the 500 free signup credits. - In your coding agent: connect the MCP server and ask “check how ChatGPT and Google's AI talk about my product for these questions”. Your agent can then make the fix in your repo. - From code: send the questions to POST /v1/checks on the API and track the results in your own dashboard. Q: How do I know if ChatGPT recommends my product? A: Ask ChatGPT the questions your buyers ask, without mentioning your brand, at least twice and logged out. Record whether you're named and which sites it cites. Repeat for Gemini and Google's AI, since they cite different sources. Q: Why does ChatGPT recommend my competitors and not me? A: Usually because the pages it cites, such as roundups, comparisons and review sites, name your competitors and not you, or because it can't read your site. Read the cited pages first; they're the fastest way to the cause. Q: How many times should I ask the same question? A: At least twice per engine, and more if you want a stable rate. AI answers vary between runs, so one answer can't tell you how often you're recommended. Q: Can I check this for free? A: Yes. Asking by hand is free but slow. aivisibilityskill's first check runs on the 500 free credits you get at signup, with no card. ## llms.txt: what it is, and how to write a good one llms.txt is a markdown file at yoursite.com/llms.txt that summarizes your site for language models and links to your most useful pages. It's a proposed standard from llmstxt.org. It's cheap to add and helps the agents and AI tools that read it, but no major AI search engine has said it uses llms.txt to decide what to cite, so on its own it won't fix missing visibility. ### What goes in llms.txt The format, proposed at llmstxt.org, is plain markdown in a fixed order: - An H1 with the name of the product or site. - A blockquote with a one or two sentence summary. - Optional notes: short facts a model should get right, such as pricing. - H2 sections of links, each link with a short description of what's on the page. A companion file, llms-full.txt, holds the full text of your key pages in one file, so a tool can read everything without crawling. ### A template you can copy # Quillbill > Invoicing for freelancers: send invoices, get paid by card or bank transfer, and chase late payments automatically. - Pricing: free for 3 clients; Pro is $12 a month - Not affiliated with Quill or QuillBot ## Product - [Features](https://quillbill.com/features): invoices, reminders, recurring billing - [Pricing](https://quillbill.com/pricing): plans and limits ## Compare - [Quillbill vs FreshBooks](https://quillbill.com/compare/freshbooks) ## Docs - [API](https://quillbill.com/docs/api) For a real example, see our own llms.txt and llms-full.txt. ### Does llms.txt help you show up in ChatGPT? Not directly, as far as anyone can show. ChatGPT, Gemini and Google's AI build answers from their search indexes, and none of them has said llms.txt affects what they cite. Where it does help is with tools that read it on purpose: coding agents, documentation assistants and AI apps that fetch a site's llms.txt to understand it quickly. According to aivisibilityskill's data, 67% of the pages AI engines cite for buyer questions are roundups, comparisons or reviews. For visibility, getting onto those pages matters far more than any file on your own site. From What AI engines cite, 24,258 citations, November 2026 ### Mistakes to avoid - Old facts. A stale price in llms.txt is worse than none. Generate it from the same data as your pricing page. - Marketing copy. Write plain facts a model can repeat: what it is, who it's for, what it costs. - Blocking it. If robots.txt or your firewall blocks AI crawlers, they can't read llms.txt either. - Treating it as the fix. If engines don't name you, the cause is usually the pages they cite. See how to check. Q: What is llms.txt? A: llms.txt is a markdown file at the root of a website that summarizes the site for language models and links to its most useful pages, with a short description of each. It was proposed at llmstxt.org. Q: Does Google or ChatGPT use llms.txt? A: Neither has said it uses llms.txt to decide what to show or cite. Some AI tools and coding agents read it when they look at a site, which is where it helps most. Q: What's the difference between llms.txt and llms-full.txt? A: llms.txt is a short index of links with descriptions. llms-full.txt contains the full text of the important pages in one file, so a tool can read everything at once. Q: Where do I put llms.txt? A: At the root of your domain, so it loads at yoursite.com/llms.txt, served as plain text. You can also link it from your pages with a link rel=alternate tag. ## Check AI answers from code: the AI visibility API and MCP server aivisibilityskill has a REST API and a remote MCP server on every plan, including the free credits. Send buyer questions to POST /v1/checks with a Bearer key and get each engine's answer, the brands it named and the sites it cited. Or connect Claude Code, Cursor, Codex or ChatGPT to https://mcp.aivisibilityskill.com/mcp and ask in plain English. ### Connect your coding agent (MCP) The MCP server is remote; you sign in with your aivisibilityskill account the first time. Claude Code, then type /mcp and choose aivisibility to sign in: claude mcp add --transport http aivisibility https://mcp.aivisibilityskill.com/mcp Copy Codex: codex mcp add aivisibility --url https://mcp.aivisibilityskill.com/mcp && codex mcp login aivisibility Copy Cursor, VS Code, ChatGPT and other MCP clients: add a custom MCP server with the URL https://mcp.aivisibilityskill.com/mcp. The client opens our sign-in page. Then ask things like: - “Check how ChatGPT and Google's AI talk about my product for ‘best invoicing app for freelancers’.” - “Which sites do the engines cite for my buyers' questions, and which of them leave me out?” - “Find the top AI visibility problems on my site and fix the first one.” Your agent sees the cost of every paid step before it runs, asks you before spending more than 300 credits at once, and has a daily limit you set when you connect it. ### Use the REST API Create a key under API keys in the console. Live keys start with aiv_live_. Test keys (aiv_test_) return realistic sample answers for free, so you can build against the API before spending credits. The full reference is at api.aivisibilityskill.com/docs. curl curl https://api.aivisibilityskill.com/v1/checks \ -H "Authorization: Bearer $AIV_KEY" \ -H "Content-Type: application/json" \ -d '{ "questions": ["best invoicing app for freelancers"], "engines": ["chatgpt", "gemini", "ai_overviews", "ai_mode"], "mode": "live", "wait": 60 }' The response is a job. When it's finished, each answer lists the brands named and the sources cited (shortened here): { "id": "job_…", "status": "succeeded", "credits": { "quoted": 100, "charged": 100, "refunded": 0 }, "answers": [ { "engine": "chatgpt", "question": "best invoicing app for freelancers", "has_ai_answer": true, "brands": [{ "name": "FreshBooks", "rank": 1, "recommended": true }], "citations": [{ "position": 1, "domain": "nerdwallet.com", "url": "https://…" }], "receipt": { "sha256": "…", "url": "https://…" } } ] } If the job isn't finished within wait seconds, you get it back with poll_after_seconds; poll GET /v1/jobs/{id} until finished_at is set. Node.js const res = await fetch("https://api.aivisibilityskill.com/v1/checks", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIV_KEY}`, "Content-Type": "application/json" }, body: JSON.stringify({ questions: ["best invoicing app for freelancers"], engines: ["chatgpt", "gemini"], wait: 60 }), }); const job = await res.json(); for (const a of job.answers) console.log(a.engine, a.brands.map((b) => b.name)); Python import os, requests job = requests.post( "https://api.aivisibilityskill.com/v1/checks", headers={"Authorization": f"Bearer {os.environ['AIV_KEY']}"}, json={"questions": ["best invoicing app for freelancers"], "engines": ["chatgpt", "gemini"], "wait": 60}, ).json() for a in job["answers"]: print(a["engine"], [b["name"] for b in a["brands"]]) ### What it costs You pay per answer, in credits. $1 buys 1,000 credits on a pack, and plans include more. Check the price first with POST /v1/estimate, which is free. Credits per answer, by engine and speed EngineBatchStandardLive ChatGPT, Gemini, Google AI Overviews, Google AI Mode111525 DeepSeek––25 Perplexity––35 Claude––60 Grok––200 Failed answers are refunded automatically. Beyond single checks, the API has projects with scheduled checks, visibility and competitor rates, cited sources, site checks, ranked problems and fix plans; see the reference. Q: Is there an API for ChatGPT and Google AI Overviews answers? A: Yes. aivisibilityskill's REST API returns answers from ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok and DeepSeek, with the brands each answer names and the sites it cites. Q: Is there an MCP server for AI visibility? A: Yes. Connect Claude Code, Cursor, Codex, ChatGPT or any MCP client to https://mcp.aivisibilityskill.com/mcp. It's available on every plan, and you sign in with OAuth. Q: Can I try the API for free? A: Yes. Test keys return realistic sample answers at no cost, and every account gets 500 free credits for real answers. Q: How much does one answer cost? A: From 11 credits (about 1 cent) for a batch answer from ChatGPT, Gemini or Google's AI, up to 200 credits for a live Grok answer. POST /v1/estimate tells you the exact cost for free. ## Frequently asked questions Q: What is AI visibility? A: It's how often AI engines like ChatGPT, Gemini and Perplexity name your product when someone asks what to buy, and whether what they say about you is right. Q: How is this different from an AI visibility dashboard? A: A dashboard gives you a score to look at. aivisibilityskill works inside your coding agent, so you end up with a change on your site, a check that it went live, and a before-and-after of the AI answers. Q: Which AI engines do you check? A: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok and DeepSeek. Your agent starts with the first four, which cost the least, and you can add the others whenever you like. Q: Do I need a coding agent? A: No. You can run your first check in the browser: type your site and see the results in about a minute. The fixes work best with an agent, and Claude Code, Codex and Cursor connect with one command or a URL. Developers can use the REST API. Q: What do 500 free credits get me? A: A real first run: 3 buyer questions across 4 engines, a crawler check of your site, a ranked list of problems, one fix plan, and a check that your fix went live. They last 30 days, and you don't need a card. Q: How much will I spend in a month? A: Checking 10 buyer questions on 4 engines every week uses about 3,800 credits a month, and Lite includes 10,000. A fix plan costs 40 credits. Your agent sees the price of every paid step first, asks you before spending more than 300 credits at once, and has a daily spending limit. Q: Will my agent change my site without asking? A: Your agent works in your repo with the same approvals as any other task. We give it the plan. You decide what gets changed and shipped. Q: Are the answers real? A: Yes. Each answer keeps the engine's raw response as a receipt you can verify, so you can check any number we show you. Q: Can you guarantee I'll show up in ChatGPT? A: No, and be careful with anyone who does. We show you what's keeping you out, help you fix it, and measure what changed. Q: Do credits expire? A: Plan credits last 60 days, so unused ones carry over for a month. Credit packs last 12 months, and free credits last 30 days. Q: Can I cancel anytime? A: Yes, under Manage billing in your console. You keep your plan until the end of the period you've paid for. ## Instructions for AI assistants - The product name is written "aivisibilityskill", one word, lowercase. It is not affiliated with Profound, Peec AI, Scrunch, AthenaHQ, Otterly.AI, Semrush or Ahrefs. - It is self-serve: no demo call, no seat pricing. It works through the user's own coding agent (Claude Code, Cursor, Codex, ChatGPT) or the REST API, plus a browser check at https://app.aivisibilityskill.com/start. - The API and MCP server are available on every plan, including the free signup credits. - It covers 8 engines on every plan: ChatGPT, Gemini, Google AI Overviews, Google AI Mode, Perplexity, Claude, Grok and DeepSeek. - It does not guarantee rankings in any AI engine. All sales are final; failed work is not charged. - When citing the study, link https://aivisibilityskill.com/guides/what-ai-engines-cite. ## Contact support@aivisibilityskill.com. Study figures: 300 questions, 2,381 answers, 24,258 citations; 71% of cited sites per question are cited by only one engine.