--- title: "AI Visibility Tracking, Ranked by Whether the Number Survives a Rerun" description: "Six ways to track your brand across AI answers, ranked by repeatability, engine coverage and whether the report leads to a fix." url: "https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis" image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/70040642-e177-43a2-b581-c4a7dc597858.png" published: "2026-10-09T05:38:13.020Z" modified: "2026-10-09T05:38:13.085Z" --- October 9, 2026 · 11 min read # AI Visibility Tracking, Ranked by Whether the Number Survives a Rerun [Alexander PerelmanHead Of Product @ groas](https://groas.com/author/alexander-perelman)[LinkedIn](https://www.linkedin.com/in/alexander-433793253/) ![Cartoon newsroom celebrates a 'LANDSLIDE!' while the whole electorate is one yawning voter: a confident result from a sample of one.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/70040642-e177-43a2-b581-c4a7dc597858.png) In this article 1. [The test: rerun the prompt before trusting the percentage](#the-test-rerun-the-prompt-before-trusting-the-percentage) 2. [6. Manual spot checks: free, honest and hard to sustain](#6-manual-spot-checks-free-honest-and-hard-to-sustain) 3. [5. SEO suites: convenient charts, shaky assumptions](#5-seo-suites-convenient-charts-shaky-assumptions) 4. [4. Standalone prompt trackers: focused, but often single-run](#4-standalone-prompt-trackers-focused-but-often-single-run) 5. [3. Rerun trackers: the first credible mention rate](#3-rerun-trackers-the-first-credible-mention-rate) 6. [2. Verified crawler logs: the check beneath the chart](#2-verified-crawler-logs-the-check-beneath-the-chart) 7. [1. Reporting with execution: the number has an owner](#1-reporting-with-execution-the-number-has-an-owner) 8. [What each number hides](#what-each-number-hides) 9. [My verdict: spend according to who will do the fixing](#my-verdict-spend-according-to-who-will-do-the-fixing) Most AI visibility scores look precise until you run the same prompt twice. If you run a business and want to know what AI answers say about you, my pick is **measurement wired to execution: repeated reporting with a log of what changed**. You get a number you can check and a record of the work behind it, without hiring an agency to babysit a spreadsheet. That is the standard for this ranking. Most AI visibility software takes one sampled answer per prompt and turns it into a confident percentage. That is a sample size of one. When SparkToro and Gumshoe ran 2,961 prompts through 600 volunteers, [under 1 in 100 repeat runs returned the same brand list](https://www.searchenginejournal.com/ai-recommendations-change-with-nearly-every-query-sparktoro/566242/), and about 1 in 1,000 kept the same order. Kevin Indig and AirOps found the same problem from the citation side: after running the same ChatGPT prompt three times across [815,000 prompt-page pairs, only 2.2% of citations remained](https://searchengineland.com/make-prompt-tracking-more-accurate-479708). Answers shift from run to run. Scores built on one answer shift with them. ## The test: rerun the prompt before trusting the percentage I used to tell clients a single ad test with 40 clicks meant something. I was wrong, and AI visibility tools are making the same mistake at higher prices. One answer per prompt is polling one person and calling it an election. Indig’s work puts [within-LLM variance at 10–34% on identical prompts](https://searchengineland.com/make-prompt-tracking-more-accurate-479708). Your Tuesday score and Wednesday score can disagree while nothing about your business has changed. I rank the six options on three checks: 1. **Repeatability:** Does the number hold up when you rerun the prompt? 2. **Engine coverage:** Can you inspect ChatGPT, Gemini, Perplexity and Google AI Overviews separately, rather than accepting one blended score? 3. **Action:** Does the result tell you what to change next? Fail the first check and the other two matter less. You are optimizing against noise. The ranking also weighs what each option is actually for: a free spot check, a dashboard, a diagnostic, or ongoing work. Not every useful tool needs to be a complete solution. It does need to be honest about what its number measures. ## 6. Manual spot checks: free, honest and hard to sustain **Who it is for:** the owner checking AI answers before spending anything. Open an incognito window and run [20 to 50 bottom-funnel buyer questions, setting `&gl=us&hl=en` for Google checks, expanding Show More, and logging whether an AI Overview appeared, whether it cited your brand and which competitors appeared](https://discoveredlabs.com/blog/google-ai-overviews-citation-tracking-how-to-measure-your-visibility). Your AI Share of Voice is then (brand citations / total Overviews triggered) × 100. It costs nothing and shows you something Google Search Console does not break out: [AI Overview performance is aggregated into Web search, with no separate click line](https://discoveredlabs.com/blog/google-ai-overviews-citation-tracking-how-to-measure-your-visibility). The one thing manual checking does better is **show you the answer, not just the score**. You can read the wording, inspect the cited pages and see which buyer questions leave you out. No blended dashboard gets between you and that result. What rules it out as ongoing measurement is the work. One pass has no reruns, no confidence interval and no history unless you become the logging system. I used to work through negative-keyword sheets at 1am; I know a manual process can work. I also know it is the first thing to slip when other work piles up. Use spot checks to choose your prompts. Stop treating them as a month-to-month score the moment someone needs to make a decision from that score. ## 5. SEO suites: convenient charts, shaky assumptions **Who it is for:** a team already paying for an SEO suite that wants AI answers in the same dashboard. The one thing suites do better is convenience. [Semrush’s AI Visibility Toolkit runs $99/month standalone per domain on 25 custom prompts at base, up to 200 on Advanced, or $199/month with Semrush One, covering ChatGPT, Perplexity, Gemini, SearchGPT and AI Overviews](https://explodingtopics.com/blog/ai-visibility-vs-brand-radar). [Ahrefs Brand Radar runs $199/month per AI platform index or $699/month for all platforms plus 2,500 custom checks](https://explodingtopics.com/blog/ai-visibility-vs-brand-radar). You get prompt counts, engine lists and history charts in a place the team already knows how to use. ![Ink cartoon of a marketer staring at a single AI answer printed as a 98% score](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/bd2cb27b-31a7-4c69-84d5-d42b6d532708.png) **What rules this tier out:** the familiar chart can make an unfamiliar measurement look settled. Rank tracking trains you to read a line as movement in a reasonably stable result. AI answers can change on the next run. If a dashboard samples once, a month-over-month line can move while your site does nothing. Prompt caps narrow the panel further; a blended cross-engine score can hide which engine changed. Use the module if convenience is the job. Skip it if you need to defend the percentage on a slide. ## 4. Standalone prompt trackers: focused, but often single-run **Who they are for:** the marketing lead who wants daily eyes on ChatGPT, Perplexity, AI Overviews and Copilot without learning an SEO suite. The one thing standalone trackers do better is focus. [Otterly Lite at $29/month ($25 on annual) covers 15 prompts across four base engines with daily tracking; Standard at $189/month covers 100 prompts plus API and Looker Studio; Premium at $489/month covers 400 prompts](https://blog.hubspot.com/marketing/hubspot-vs-otterly). The workflow is simple: prompts in, mentions out. For a small brand list, that is enough to start seeing what gets named. What rules this tier out for a defensible score is that **one daily answer is still one answer**. Daily tracking gives you more dates, not necessarily repeat runs of the same prompt on each date. A changed mention could be a real shift or just another answer to the same question. Coverage can be narrower than the engine list you want, too: [Gemini, Claude and AI Mode sit behind paid add-ons on every Otterly tier](https://blog.hubspot.com/marketing/hubspot-vs-otterly). Buy this tier to keep watch. Do not confuse that watchlist with a repeatable appearance rate. ![Bar chart showing the same AI prompt cited only twice across ten reruns](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/bf1c9523-888d-4ad3-b930-4c890067aa58.png) ## 3. Rerun trackers: the first credible mention rate **Who they are for:** anyone whose report has to survive contact with Wednesday. The thing this approach does better is treat prompt tracking like polling rather than rank tracking. One workable method is [40 seed prompts split across brand, category and problem terms, with five repetitions per prompt per engine each week, tracked separately by engine as a mention rate with a confidence interval](https://searchengineland.com/make-prompt-tracking-more-accurate-479708). The mechanism matters. Repeats turn one shaky answer into an appearance rate. Per-engine splits stop a strong Perplexity showing from concealing a weak ChatGPT showing. The interval gives you a way to ask whether a six-point move looks like a change or sits inside the usual spread. Keep the panel fixed, or a new mix of prompts can move the score even when the answers to the old prompts have not improved. I laid out the panel-building method in [how to measure AI search visibility without trusting one screenshot](https://groas.com/post/how-to-measure-your-brand-s-visibility-i-2). **What rules this tier out as a complete service:** it stops at measurement. A credible gap is useful if someone owns the next step. If nobody does, you have paid for a better description of the problem. Still, if your report goes to a boss, this is the first prompt-tracking tier I would trust over a single-run chart. Get the repeated number before debating the color of the dashboard. ## 2. Verified crawler logs: the check beneath the chart **Who they are for:** the operator who suspects the pretty chart is lying. Prompt trackers tell you what answers said. Logs tell you whether the engines fetched your pages at all. That is not the same as being cited, but it is a useful check when a citation rate stalls and you need to investigate why. The one thing logs do better is **show verified fetching instead of inferred visibility**. The mechanism matters because [tag analytics see zero AI crawler traffic: GPTBot and OAI-SearchBot do not run JavaScript](https://kitbase.dev/blog/gptbot-explained). Look only at normal tag-based analytics and you can miss crawler activity. Server or CDN logs are where you inspect it. Do not count every request carrying a bot name. A user-agent can be spoofed, so raw `GPTBot` hit counts are not proof until you [verify the IP against openai.com/gptbot.json, searchbot.json and chatgpt-user.json, or check the Web Bot Auth signature](https://kitbase.dev/blog/gptbot-explained). Then distinguish the bots: [blocking GPTBot opts out of training crawls, not ChatGPT citations; OAI-SearchBot drives search citations, while ChatGPT-User fetches bypass robots.txt as user-triggered](https://kitbase.dev/blog/gptbot-explained). Otherwise, it is easy to block one bot, watch spoofed hits accumulate and draw the wrong conclusion about visibility. What rules logs out as a standalone score is simple: **a fetch is not a mention**. Logs pass a different repeatability test from prompt tracking. You can verify whether requests happened, but you cannot calculate an answer mention rate from them. I rank them above rerun trackers as a diagnostic that can check the story behind the chart, not as a replacement for tier 3. Pair the two when you need to know both what the answer said and whether your pages were being fetched. ## 1. Reporting with execution: the number has an owner **Who it is for:** the owner who does not want a second subscription just to find out what needs fixing. The mechanism is unglamorous: the repeated panel from tier 3, per-engine mention rates, a check on whether engines fetch your pages, and work on pages, structure and citations when the rate stalls. The action log belongs next to the measurement. You should be able to see what was rewritten, what was fixed technically and why, rather than receiving a percentage and a separate suggestion to do something about it. The one thing this option does better is **connect the observed change to work someone actually did**. It does not make every move in the mention rate attributable to one edit; AI answers still vary. It does give you a record you can inspect alongside the repeated numbers. If you are weighing whether to pay for that second half at all, I made the case for and against it in [is AI prompt tracking worth paying for](https://groas.com/post/is-tracking-ai-prompts-worth-paying-for). What rules it out: you only need a number for a quarterly deck and nobody will act on it. You would pay for execution you will not use. Tiers 4 through 6 fail my repeatability test when they rely on a single run. Tier 3 passes but stops at the chart. [Standalone trackers stop at recommendations by design: you get the gap, and the fixing happens outside the tool](https://blog.hubspot.com/marketing/hubspot-vs-otterly). That is fine if you have someone to do the fixing. If you do not, the report risks becoming a PDF nobody opens twice. The version I run for clients is [groas Earned Search](https://groas.com/for-businesses). I would put a budget behind the work and its measurement together, not a confident percentage on its own. ## What each number hides Every reporting tier has a way to make a thin sample look substantial. Manual checks hide it behind effort: you remember the prompts you ran and forget the reruns you did not. Suites hide it behind design: a line chart that looks like rank tracking feels stable even when the answers underneath have changed. Single-run trackers hide it behind coverage: four engine logos at the top of a dashboard do not mean each percentage has earned your confidence. Logs hide something different. They prove fetching happened and invite you to treat fetching as citing. It is not. **If a prompt report cannot show repeated runs and the spread around its average, the percentage on page one is decoration.** Ask to see the runs before you ask whether the dashboard exports to a deck. ![Paper craft layers showing reporting stacked above fixing work](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/57bae811-790e-4156-a187-c57523329ec8.png) ## My verdict: spend according to who will do the fixing Under $100/month, keep it small and repeated. Run a manual 20-to-50-prompt panel monthly for calibration, and put one cheap tracker on daily duty for the prompts that carry buying intent. Say you spend $20k/month on search: 15 prompts on an entry plan can cover brand terms, five category prompts and five problem prompts, with room to rotate. The tracker keeps watch; your manual reruns check whether a movement deserves attention. Do not present its single daily sample as a settled monthly rate. Skip the suite module unless you already pay for the suite, and skip logs until someone on your side can read them. | Option | Survives a rerun | Split by engine | Tells you the fix | | --------------------------- | -------------------------- | ------------------------- | ----------------- | | 6. Manual checks | No, unless you repeat them | Yes, if you log it | No | | 5. Suite AI module | Not on a single-run score | May be blended | No | | 4. Single-run tracker | No | Partly | No | | 3. Rerun tracker | Yes | Yes | No | | 2. Verified logs | Yes, for fetching | N/A: fetches, not answers | Partly | | 1. Reporting with execution | Yes | Yes | Yes | Done-for-you starts to earn its place when the fixing has no owner. A tracker can flag a missing citation, but someone still has to rewrite the page, address the technical block or pursue the third-party mention. That person bills by the hour or never gets to it. I left one popular option off the ranking on purpose: the _one-time AI audit PDF_. It scores you once across fifty prompts, hands you forty pages and never reruns anything. It fails the test and leaves you holding the work. For the reader who skipped to the end: **buy measurement that reruns, splits by engine and ships with the changes. Otherwise, keep your money and keep the spreadsheet.** ## Frequently Asked Questions ### Why do AI visibility scores change when nothing about my business has changed? Most AI visibility tools base their score on one sampled answer per prompt, and AI answers differ from run to run. SparkToro and Gumshoe found under 1 in 100 repeat runs returned the same brand list, and Kevin Indig and AirOps measured within-LLM variance of 10–34% on identical prompts. A score built on a single answer shifts along with the answers. ### What should I look for when choosing an AI visibility tracking tool? Judge each option on three checks: repeatability, meaning the number holds up when you rerun the prompt; engine coverage, meaning you can inspect ChatGPT, Gemini, Perplexity and Google AI Overviews separately; and action, meaning the result tells you what to change next. Fail repeatability and the other two matter less, because you are optimizing against noise. ### Can I just manually check what ChatGPT says about my brand instead of paying for a tool? Manual spot checks cost nothing and let you read the actual wording, inspect cited pages and see which buyer questions leave you out. Run 20 to 50 bottom-funnel buyer questions in an incognito window and log whether an AI Overview appeared, whether it cited your brand and which competitors appeared. They work for choosing prompts and calibration, but become unreliable as a month-to-month score because one pass has no reruns, no confidence interval and no history. ### Is Semrush's AI Visibility Toolkit reliable for tracking AI answers? Semrush's AI Visibility Toolkit runs $99/month standalone per domain on 25 custom prompts, up to 200 on Advanced, or $199/month with Semrush One, covering ChatGPT, Perplexity, Gemini, SearchGPT and AI Overviews. It is convenient if your team already uses Semrush, but a familiar chart can make an unfamiliar measurement look settled: if the dashboard samples once, a month-over-month line can move while your site does nothing. Use it for convenience, not to defend a percentage. ### Does daily tracking in a standalone tool like Otterly give a reliable mention rate? No. Daily tracking gives you more dates, not necessarily repeat runs of the same prompt on each date, so one daily answer is still one answer. A changed mention could be a real shift or just another answer to the same question. Otterly Lite at $29/month covers 15 prompts across four base engines, and Gemini, Claude and AI Mode sit behind paid add-ons on every tier, so coverage can be narrower than the engine list suggests. ### How do you measure AI visibility so the number is repeatable? Treat prompt tracking like polling: run a fixed panel of 40 seed prompts split across brand, category and problem terms, with five repetitions per prompt per engine each week, and track a mention rate separately by engine with a confidence interval. Repeats turn one shaky answer into an appearance rate, per-engine splits stop a strong Perplexity showing from hiding a weak ChatGPT one, and the interval shows whether a move is real or usual spread. Keep the panel fixed so new prompts do not move the score. ### Do AI crawler hits in my analytics prove my pages are being cited? No, a fetch is not a mention. Tag-based analytics often see zero AI crawler traffic because GPTBot and OAI-SearchBot do not run JavaScript, so check server or CDN logs instead. Do not count every request carrying a bot name, since user-agents can be spoofed; verify the IP against openai.com/gptbot.json, searchbot.json and chatgpt-user.json, or check the Web Bot Auth signature. Also note that blocking GPTBot opts out of training crawls, not ChatGPT citations, while OAI-SearchBot drives search citations. ### Are one-time AI audit reports worth paying for? The article leaves the one-time AI audit PDF off its ranking on purpose: it scores you once across fifty prompts, hands you forty pages and never reruns anything, so it fails the repeatability test and leaves you holding the work. Buy measurement that reruns, splits by engine and ships with the changes instead, or keep your money and keep the spreadsheet. ## Pay For Results, Not For Hours Businesses buy the outcome, agencies resell it, and groas answers for it either way. [See If You Qualify](https://groas.typeform.com/to/xC1bQNUT) ## Structured data ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#webpage","url":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis","name":"AI Visibility Tracking, Ranked by Whether the Number Survives a Rerun","description":"Six ways to track your brand across AI answers, ranked by repeatability, engine coverage and whether the report leads to a fix.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/70040642-e177-43a2-b581-c4a7dc597858.png"},"breadcrumb":{"@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#breadcrumb"},"dateModified":"2026-10-09T05:38:13.085Z","mainEntity":{"@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"AI Visibility Tracking, Ranked by Whether the Number Survives a Rerun","item":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#article","url":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis","mainEntityOfPage":{"@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"AI Visibility Tracking, Ranked by Whether the Number Survives a Rerun","description":"Six ways to track your brand across AI answers, ranked by repeatability, engine coverage and whether the report leads to a fix.","datePublished":"2026-10-09T05:38:13.020Z","dateModified":"2026-10-09T05:38:13.085Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/70040642-e177-43a2-b581-c4a7dc597858.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"whether, prompt, visibility, tracking, number, rerun, answers, prompts","wordCount":2227,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"Ex Goldman Sachs and Ex Stanford Computer Science","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/every-way-to-measure-your-brand-s-ai-vis#faq","mainEntity":[{"@type":"Question","name":"Why do AI visibility scores change when nothing about my business has changed?","acceptedAnswer":{"@type":"Answer","text":"Most AI visibility tools base their score on one sampled answer per prompt, and AI answers differ from run to run. SparkToro and Gumshoe found under 1 in 100 repeat runs returned the same brand list, and Kevin Indig and AirOps measured within-LLM variance of 10–34% on identical prompts. A score built on a single answer shifts along with the answers."}},{"@type":"Question","name":"What should I look for when choosing an AI visibility tracking tool?","acceptedAnswer":{"@type":"Answer","text":"Judge each option on three checks: repeatability, meaning the number holds up when you rerun the prompt; engine coverage, meaning you can inspect ChatGPT, Gemini, Perplexity and Google AI Overviews separately; and action, meaning the result tells you what to change next. Fail repeatability and the other two matter less, because you are optimizing against noise."}},{"@type":"Question","name":"Can I just manually check what ChatGPT says about my brand instead of paying for a tool?","acceptedAnswer":{"@type":"Answer","text":"Manual spot checks cost nothing and let you read the actual wording, inspect cited pages and see which buyer questions leave you out. Run 20 to 50 bottom-funnel buyer questions in an incognito window and log whether an AI Overview appeared, whether it cited your brand and which competitors appeared. They work for choosing prompts and calibration, but become unreliable as a month-to-month score because one pass has no reruns, no confidence interval and no history."}},{"@type":"Question","name":"Is Semrush's AI Visibility Toolkit reliable for tracking AI answers?","acceptedAnswer":{"@type":"Answer","text":"Semrush's AI Visibility Toolkit runs $99/month standalone per domain on 25 custom prompts, up to 200 on Advanced, or $199/month with Semrush One, covering ChatGPT, Perplexity, Gemini, SearchGPT and AI Overviews. It is convenient if your team already uses Semrush, but a familiar chart can make an unfamiliar measurement look settled: if the dashboard samples once, a month-over-month line can move while your site does nothing. Use it for convenience, not to defend a percentage."}},{"@type":"Question","name":"Does daily tracking in a standalone tool like Otterly give a reliable mention rate?","acceptedAnswer":{"@type":"Answer","text":"No. Daily tracking gives you more dates, not necessarily repeat runs of the same prompt on each date, so one daily answer is still one answer. A changed mention could be a real shift or just another answer to the same question. Otterly Lite at $29/month covers 15 prompts across four base engines, and Gemini, Claude and AI Mode sit behind paid add-ons on every tier, so coverage can be narrower than the engine list suggests."}},{"@type":"Question","name":"How do you measure AI visibility so the number is repeatable?","acceptedAnswer":{"@type":"Answer","text":"Treat prompt tracking like polling: run a fixed panel of 40 seed prompts split across brand, category and problem terms, with five repetitions per prompt per engine each week, and track a mention rate separately by engine with a confidence interval. Repeats turn one shaky answer into an appearance rate, per-engine splits stop a strong Perplexity showing from hiding a weak ChatGPT one, and the interval shows whether a move is real or usual spread. Keep the panel fixed so new prompts do not move the score."}},{"@type":"Question","name":"Do AI crawler hits in my analytics prove my pages are being cited?","acceptedAnswer":{"@type":"Answer","text":"No, a fetch is not a mention. Tag-based analytics often see zero AI crawler traffic because GPTBot and OAI-SearchBot do not run JavaScript, so check server or CDN logs instead. Do not count every request carrying a bot name, since user-agents can be spoofed; verify the IP against openai.com/gptbot.json, searchbot.json and chatgpt-user.json, or check the Web Bot Auth signature. Also note that blocking GPTBot opts out of training crawls, not ChatGPT citations, while OAI-SearchBot drives search citations."}},{"@type":"Question","name":"Are one-time AI audit reports worth paying for?","acceptedAnswer":{"@type":"Answer","text":"The article leaves the one-time AI audit PDF off its ranking on purpose: it scores you once across fifty prompts, hands you forty pages and never reruns anything, so it fails the repeatability test and leaves you holding the work. Buy measurement that reruns, splits by engine and ships with the changes instead, or keep your money and keep the spreadsheet."}}]}]} ```