DEV Community: AEOGEOAI Research Team

The Audience of One Problem: Why You Can't Trust Your Own AI Visibility Test

AEOGEOAI Research Team — Mon, 13 Jul 2026 02:08:29 +0000

When you ask an LLM whether it recommends your product, you're running a test where you're both the subject and the observer, from inside the system you're measuring. Here's a clean-room protocol to test it properly.

Canonical URL: https://aeogeoai.net/blog-why-checking-your-own-ai-visibility-lies

You want to know whether ChatGPT recommends your product. So you open it and ask: "What's the best tool for X?" Your product shows up. You close the tab, reassured.

You just ran a broken experiment.

Not because the answer was wrong, but because of how you got it. You queried a system you're trying to measure, from inside that system, using an identity it already associates with the thing you're measuring. Every variable that should have been controlled was contaminated. If a colleague showed you that methodology in a code review, you'd reject it.

This is the Audience of One problem, and it catches almost everyone who tries to check their own AI visibility.

Why your own session is a contaminated test environment

LLM-backed search doesn't return one universal answer. Public documentation from the major providers indicates outputs can be shaped by signals about the request: location, conversation history, personalization settings, account state, and the model version in use. That's fine — until you try to use your own session as a measurement instrument.

When you check your own product, you introduce confounding variables you'd never accept in any other test:

Observer identity: you're often signed in to an account that has searched for your own product before.
Location bias: you're querying from the same network and region your business operates in.
Session history: prior turns in the same conversation prime the model toward your brand.
Selection effect: you phrase the query the way you think about your product, not the way a stranger would.

Each of these nudges the result toward you. The output feels objective — it's a fresh-looking answer on a clean-looking screen — but it's the single most flattering measurement you could possibly take. You are an audience of one, and the test was built to please you.

The tell: ask the model directly, "Would you recommend my product?" and it will often say yes enthusiastically. That's not evidence. A model can't reliably predict its own future retrievals, and after a conversation about your product it's biased toward agreeing with you. Willingness is not behavior.

The clean-room protocol

The fix is the same one you'd apply to any contaminated experiment: remove the confounds and reproduce. Here's the procedure.

1. Strip your identity

Log out completely, or use a temporary/incognito session with personalization/memory disabled (private browsing alone doesn't remove account or location signals).
Best case: run it on a device and network that has never searched for your product. A colleague's phone on cellular data is a surprisingly good clean room.

2. Query as a stranger would

Ask the actual user's question — "best [category] tool for [use case]" — not your brand name.
Do not mention your product until after the tool has answered. Naming it first re-contaminates the session.
Phrase it the way a real user would, not the way your marketing does.

3. Read the source list, not the prose

Use a tool that exposes its citations (Perplexity is good for this — it shows which domains it actually pulled).
You're checking whether your domain appears in the retrieved sources, not whether the model says something nice. Retrieval is the ground truth; the prose is downstream of it.

4. Reproduce across sessions and engines

Run it several times. Retrieval varies per session — one run proves almost nothing.
Run it across ChatGPT, Claude, and Gemini. They retrieve differently; being cited by one is not being cited by all.
Log what appeared each time. A stable pattern is signal; a single lucky hit is noise.

5. Interpret with tiers

Keep three buckets separate, because conflating them is how people fool themselves:

Observed: what actually appeared in the retrieved sources across clean runs.
Documented: what providers publicly say influences outputs.
Inferred: your theory about why — held loosely, not stated as fact.

If your product shows up in the clean, logged-out, reproduced, third-party test — that's real. If it only appears when you check from your own chair, you've been measuring an audience of one.

Why this matters beyond vanity

If you're building anything that depends on being discoverable through AI answers — a dev tool, a SaaS, an API — your own perception of your visibility is the least reliable data point you have. The contaminated test doesn't just flatter you; it hides the gap between what you see and what a prospective user sees. And the prospective user is the only observer who matters.

We ran this on our own site. The signed-in check looked healthy. The clean-room test disagreed: for the query that mattered most we weren't retrieved at all, and where we did appear, the engine had surfaced the wrong page. None of that was visible from the comfortable version of the test. We only found it by running the uncomfortable one.

Build the clean room. Test from outside your own reflection. It's the only version of the result that isn't lying to you.

I research AI search visibility and citation behavior at AEOGeoAI. If you want to see the multi-engine version of this test run automatically, there's a free checker on the site — the manual protocol above costs nothing and teaches you more.

Why Your 150kb WordPress Header Is Making You Invisible to AI Crawlers

AEOGEOAI Research Team — Sat, 04 Jul 2026 10:29:48 +0000

AI retrieval systems do not read pages the way humans do.

A human scrolls past the navigation, ignores the cookie banner, skips the breadcrumb trail, and reads the article. An AI crawler processes the HTML source in the order it appears — and decides whether the page is worth citing before it has finished reading.

If your opening content block is dominated by structural noise rather than direct answers, your Answer Density is low. Low Answer Density means low citation probability, regardless of how good the actual article is.

What HTML Noise Actually Is

HTML Noise is the structural content that appears before your article body in raw HTML source — navigation menus, promotional banners, breadcrumbs, author metadata, affiliate disclosures, social sharing buttons, cookie consent notices, tables of contents.

On a lightweight custom site, this might be 2–3kb before the first paragraph. On a feature-heavy WordPress installation with a popular theme and a stack of plugins, it can easily be 40–80kb before the AI crawler reaches the first sentence of your actual content.

Here is what a low-Answer-Density WordPress page looks like in raw source:

<!-- Navigation: 800 tokens -->
<nav class="primary-nav">
  <div class="nav-wrapper">
    <div class="logo-container">...</div>
    <ul class="menu-items">
      <li><a href="/about">About</a></li>
      <!-- 20+ more menu items -->
    </ul>
    <div class="mobile-toggle">...</div>
  </div>
</nav>

<!-- Hero banner -->
<div class="hero-banner promo-banner sticky">
  <p>🔥 Summer sale — 40% off all plans. Use code SUMMER40</p>
</div>

<!-- Breadcrumb -->
<div class="breadcrumb-wrapper">
  <span>Home</span> › <span>Blog</span> › <span>Category</span> › <span>Article</span>
</div>

<!-- Author block -->
<div class="author-meta">
  <img src="avatar.jpg" class="author-avatar" />
  <div class="author-info">
    <span class="author-name">Jane Smith</span>
    <span class="author-title">Content Writer at [Company]</span>
    <span class="publish-date">June 12, 2026 · 8 min read</span>
  </div>
</div>

<!-- Table of contents (injected by plugin) -->
<div class="toc-container">
  <p>In this article:</p>
  <ul>
    <li><a href="#section1">Introduction</a></li>
    <!-- more links -->
  </ul>
</div>

<!-- Affiliate disclosure -->
<div class="disclosure-box">
  This article contains affiliate links. We may earn a commission...
</div>

<!-- ACTUAL CONTENT STARTS HERE — after 3,000+ words of structural noise -->
<h1>How to Choose the Best CRM Software</h1>
<p>Choosing the right CRM can be difficult...</p>

An AI retrieval system evaluating this page has processed thousands of tokens of navigation, promotions, metadata, and disclosures before it reaches the article heading. The relevance signal it calculates for this page is based on the aggregate of everything above — and that aggregate is structurally noisy.

What High Answer Density Looks Like

Answer Density is how much of a retrievable content block directly answers the question being asked. High Answer Density means the opening section contains a direct, specific answer with named entities. Low Answer Density means the opening section contains introductions, context, or background without substantive answers.

Compare these two openings for an article titled "How to Choose the Best CRM Software":

Low Answer Density:

There are many factors to consider when choosing a CRM. In this article, we'll walk through the most important things to think about before making a decision.

High Answer Density:

The three CRM platforms most frequently recommended for small sales teams are HubSpot (free tier, strong contact management), Pipedrive (pipeline-first, simple UI), and Close (built specifically for outbound sales). The right choice depends on whether your team's primary workflow is inbound lead nurturing, outbound prospecting, or account management.

The second version names entities, makes specific claims, and provides a direct answer in the first two sentences. AI systems decide a page is worth citing largely on the basis of the first content block — before processing the rest of the page.

The WordPress-Specific Problem

WordPress makes HTML Noise worse by default because:

Themes inject structural elements before <article> — navigation, sidebars, hero sections, breadcrumbs
Plugins add content to the page body — TOC plugins, social share bars, affiliate disclosure boxes, cookie consent injectors
The Gutenberg block model adds wrapper divs — each block adds at least one <div> wrapper with class attributes before the actual content node

A 2026 audit of WordPress sites using popular themes (Divi, Avada, OceanWP) found average HTML output of 120–180kb before the first <h1> tag on article pages. A lightweight static HTML equivalent would achieve the same first heading in under 5kb.

The Practical Fixes

1. Switch to a lightweight theme

Themes with minimal HTML output: GeneratePress, Kadence, Blocksy, Astra (with minimal blocks). These produce 8–15kb of HTML before content on a clean install, compared to 60–150kb for feature-heavy themes.

2. Move social share buttons below content

Most social share plugins inject markup before the article. Move them to after_content hooks or below the closing </article> tag.

// functions.php — move social share to after content
remove_action( 'the_content', 'social_share_before_content', 5 );
add_action( 'the_content', 'social_share_after_content', 25 );

3. Disable breadcrumbs on article pages

Breadcrumbs provide navigation value but carry no content signal. On article templates, disable them:

// Disable Yoast breadcrumbs on single posts
add_filter( 'wpseo_breadcrumb_output', function( $output ) {
  if ( is_single() ) return '';
  return $output;
});

4. Front-load your opening paragraph

Regardless of theme, make the first <p> tag after your <h1> a direct answer. The pattern that produces the highest Answer Density:

[Named entity] is [category definition].
[Direct answer to the question the title implies].
[Specific claim with data or named examples].

Do not start with "In this article we will..." — that is pure HTML Noise in paragraph form.

5. Conditionally delay non-essential script injections

Scripts that inject content before </head> delay HTML parsing. Move non-essential scripts to defer or async:

<!-- Before: blocks parsing -->
<script src="tracking-pixel.js"></script>

<!-- After: deferred, does not block content parsing -->
<script src="tracking-pixel.js" defer></script>

For Cloudflare deployments, use Rocket Loader to defer non-critical JS automatically:

Cloudflare Dashboard → Speed → Optimization → Rocket Loader → On

Measuring Your Answer Density

A quick manual audit: right-click any page → View Page Source → search for your article's <h1> tag. Count how many lines of HTML appear above it. On a well-optimised page: under 100 lines. On a typical WordPress install with a feature theme: 300–800 lines.

The ratio of pre-content HTML to content HTML is a rough proxy for HTML Noise density.

The Eligibility Question Underneath

HTML Noise and Answer Density affect how AI systems evaluate your content once they can access it. But there is a more fundamental question: can AI systems access your content at all?

A site can have perfect Answer Density and still score zero in ChatGPT answers if OAI-SearchBot is blocked by Cloudflare's Block AI Bots setting, or if the site was never submitted to Bing Webmaster Tools (which ChatGPT uses as its primary index).

The distinction between AI Search Eligibility (can the system see you?) and AI Search Visibility (does the system mention you?) is the frame that makes troubleshooting systematic rather than guesswork. Fix eligibility first, then Answer Density, then third-party citation coverage.

About the Author

I build open diagnostic utilities focused on AI search visibility and entity citation analysis. If you want to check how your site's content is interpreted across ChatGPT, Claude, and Gemini simultaneously, run a free check at aeogeoai.net — it returns 0–100 scores per model and word-for-word excerpts showing exactly what each AI system says about your brand. Three free checks, no account required.

Stop Optimizing for Backlinks: The Math Behind the AI Recommendation Formula

AEOGEOAI Research Team — Mon, 29 Jun 2026 09:44:15 +0000

Two decades of SEO logic said backlinks were the primary trust signal. More links from authoritative domains meant higher rankings. The entire link-building industry was built on this assumption.

AI search inverts it.

Brand mentions — independent third-party references to your entity — correlate 3x more strongly with AI citation than backlinks: 0.664 vs 0.218 correlation coefficient (Goralewicz, Onely 2026). The signal that built Google rankings is not the signal that gets you into ChatGPT answers.

Understanding why requires a framework: the AI Recommendation Formula

The Formula

AI Recommendation Probability =
Citation Coverage
× Category Clarity
× Review Presence
× Third-Party Authority
× Evidence Consistency

Five factors. Multiplication — not addition.

This is the critical mathematical property: a near-zero score on any single factor collapses the entire output regardless of the strength of the others. A brand with dominant category clarity, strong reviews, and excellent third-party authority still scores near zero if it has almost no citation coverage. The multiplication means there is no compensation across factors.

Factor 1: Citation Coverage

The volume of independent indexed mentions of your entity across sources AI models weight heavily — editorial publications, industry directories, Reddit discussions, comparison articles, G2, Capterra, community forums.

This is the factor most people underinvest in because it is not their own property. It requires other people and publications to write about you. For most independent businesses and newer brands, this is structurally absent — they exist only on their own website.

The AEOGeoAI New Jersey study tested 216 independent health practices across ChatGPT, Claude, and Gemini. 98% scored zero across all three models. The primary cause: near-complete absence of third-party citation coverage. A dentist in Hackensack with 200 five-star Google reviews and a fully optimised website had almost no indexed third-party coverage that AI systems could cross-reference.

Google reviews do not count. They are first-party data on Google's own platform. AI systems need independent sources.

Factor 2: Category Clarity

How consistently your entity is described in the same category terms across all sources AI learns from.

A brand described as "project management software" on its own site, "team collaboration tool" on G2, "productivity platform" on Reddit, and "workflow automation app" on a review site has low category clarity. AI systems aggregate these descriptions and may produce a blended, inaccurate, or uncertain category assignment.

Category clarity is damaged by inconsistency and repaired by repetition. Every source that describes your brand in the same terms reinforces the category signal. This is why the Brand Definition Formula matters:

[Brand] is a [category] for [audience].
It does [core function].
It solves [specific problem].

This exact formulation, repeated verbatim across your own content, press releases, guest posts, and product listings, strengthens AI category assignment. Variation hurts.

Factor 3: Review Presence

Structured presence on authoritative review platforms with accurate, current descriptions.

Not review volume — review presence. AI systems use review platforms as authoritative structured sources for entity data. A brand present on G2, Capterra, Trustpilot, or relevant vertical directories with accurate category descriptions gains a citation source that AI retrieval systems treat as reliable.

The key detail: the description on the review platform matters as much as the rating. A G2 profile that describes your software incorrectly contributes to the Evidence Problem regardless of your star rating.

Factor 4: Third-Party Authority

The quality and domain authority of the external sources discussing your brand.

A mention in a 15-year-old industry trade publication carries more weight than a mention in a two-week-old Medium post. An editorial article on a national news site (MSN, Forbes, industry verticals) produces a different signal than a user-generated forum comment. AI systems are not naive about source quality — they apply something close to what traditional SEO called domain authority to citation sources.

This is why structured placement on high-authority publications produces faster citation results than organic mentions on low-authority sources. The authority of the citing source is a multiplier on the citation value.

Factor 5: Evidence Consistency

The degree to which key entity facts are described identically across all sources.

Brand name, website URL, founding year, primary service, location, founding team — these facts, if they appear inconsistently across sources, create what AI systems experience as conflicting evidence. Conflicting evidence reduces citation confidence. AI systems prefer to recommend entities they can describe accurately. If they cannot produce a consistent description, they tend to omit the entity.

This is the AI equivalent of NAP consistency in local SEO, but broader: it applies to every factual claim about your entity across every indexed source.

The Evidence Gap

The Evidence Gap describes the difference between what AI systems know about your brand and what they know about your competitors.

Most brands that score zero in AI search do not have a product problem. They have an Evidence Problem: AI systems have insufficient, inconsistent, or absent evidence to include the brand confidently in recommendations.

Diagnosing the Evidence Gap means auditing each of the five factors:

Citation Coverage: How many independent indexed sources mention you?
Category Clarity: Is your category description consistent across sources?
Review Presence: Are you on the platforms AI uses as structured data sources?
Third-Party Authority: What is the domain quality of your citing sources?
Evidence Consistency: Do your key facts match across all sources?

A brand scoring weak on Citation Coverage and Evidence Consistency with strong Review Presence and Category Clarity can prioritise precisely: acquire more independent citations and audit factual consistency before touching anything else.

What This Means for Development Teams

The practical implication for developers building brand presence:

Your own website is one source. AI systems aggregate many sources. Investing entirely in on-site optimisation — schema, content, page speed, structured data — produces diminishing returns for AI visibility because you are only ever improving one citation source.

Schema markup helps AI understand your content. It does not create the third-party citation coverage that drives the Citation Coverage factor. These are separate investments targeting separate mechanisms.

The brands that score 80+ in AI visibility checks consistently have strong performance across all five factors — not perfection in two.

Measuring Your Formula Score

You can run a quick citation coverage audit right now:

Search "[your brand name]" (in quotes) on Google — count the unique domains referencing you
Search your brand on G2, Capterra, Trustpilot — are you listed with accurate descriptions?
Search site:reddit.com "[your brand]" — are you mentioned in community discussions?
Check Wikipedia, Crunchbase, LinkedIn, industry directories — do you appear?

Then compare the same audit for your top competitor. The gap between your count and theirs is roughly your Evidence Gap.

To check how that translates to actual AI recommendation probability, run a free check at aeogeoai.net — it returns 0–100 scores per model (ChatGPT, Claude, Gemini) and the exact text each AI system uses to describe your brand. Three free checks, no account required.

About the Author

I work in Miami Beach and I build open network diagnostic utilities focused on AI search visibility and entity citation analysis. If you want to test your brand's AI Recommendation Formula score across ChatGPT, Claude, and Gemini simultaneously, run a free check at aeogeoai.net.

The ChatGPT Invisibility Bug: Why High-Quality Content Fails to Index in LLM Search

AEOGEOAI Research Team — Sat, 27 Jun 2026 07:04:42 +0000

You built a fast site. Clean HTML. Proper schema. Good content. You checked your Google Search Console — indexed, ranking, healthy.

Then someone tells you ChatGPT has no idea your site exists.

This is not a content problem. It is an eligibility problem — and most developers confuse the two.

Eligibility vs Visibility

Before your site can appear in AI-generated answers, it needs to satisfy two completely separate conditions:

AI Search Eligibility is whether an AI system's retrieval infrastructure can access, crawl, and index your content at all. It is a binary gate.
AI Search Visibility is how prominently your brand appears once that gate is open. It is a $0\text{–}100$ spectrum.

Most content and SEO advice talks strictly about visibility — schema, structured content, brand mentions, third-party citations. That advice is irrelevant if your site never clears the eligibility gate. A site can have perfect structured data, excellent backlinks, and a strong content strategy, and still score zero in ChatGPT Search if it has never been submitted to Bing Webmaster Tools.

The Bing Connection Most Developers Miss

ChatGPT Search is not powered by Google. It uses Microsoft Bing as its primary index.

This means your Google Search Console setup, your Googlebot permissions, your Google-verified sitemap — none of it makes your content eligible for ChatGPT. You need a parallel Bing infrastructure:

Bing Webmaster Tools — verify your site and submit your sitemap
OAI-SearchBot — OpenAI's crawler must not be blocked
IndexNow — push URL updates directly to Bing in real time

If any of these are missing, ChatGPT Search cannot see your content regardless of quality. The same applies to Microsoft Copilot, Bing AI Mode, and Microsoft 365 Copilot — they all run on the same index. One submission, four AI surfaces.

The Cloudflare "Block AI Bots" Problem

Here is where it gets specific to the Cloudflare stack.

Cloudflare's Block AI Bots security setting — found under Security → Bots — is designed to block AI training crawlers. But its wildcard implementation also blocks OAI-SearchBot, the crawler that feeds ChatGPT Search results.

If you enabled Block AI Bots and never checked the fine-grained rules, you may have disabled ChatGPT Search eligibility for your entire site without realizing it. Check your Cloudflare settings now under Security → Bots → Bot Fight Mode / Block AI Bots. If it is enabled, you have two options:

Disable it entirely (simplest)
Create a WAF custom rule to allow OAI-SearchBot by user agent before the block rule fires

# Cloudflare WAF — allow OAI-SearchBot before AI block rule
(http.user_agent contains "OAI-SearchBot") → Allow

The same applies to PerplexityBot if you want Perplexity eligibility, and Google-Extended if you want to appear in Google AI training data.

The robots.txt Wildcard Trap

Beyond Cloudflare, check your robots.txt for wildcard disallow rules:

User-agent: *
Disallow: /api/
Disallow: /admin/

A wildcard User-agent: * applies to every bot not explicitly listed elsewhere in the file. If you have not added explicit Allow rules for AI crawlers, your wildcard rules may be blocking them.

The fix is explicit permissions:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: *
Disallow: /api/
Disallow: /admin/

Note: List AI crawlers before your wildcard block. Order matters in robots.txt.

IndexNow: Pushing Updates to Bing in Real Time

Once your site is eligible, freshness matters. 76.4% of ChatGPT's most-cited pages were updated within the last 30 days. A site that submits content changes immediately has a structural advantage over one that waits for Bing's crawl cycle.

IndexNow is a protocol that pushes URL change notifications directly to Bing (and Yandex) the moment content updates. Cloudflare supports it natively via Crawler Hints:

Cloudflare Dashboard → Speed → Optimization → Crawler Hints → Enable

With Crawler Hints enabled, Cloudflare automatically notifies Bing via IndexNow whenever a page is updated. No plugin, no API calls, no scheduled jobs. For non-Cloudflare stacks, the IndexNow API call is straightforward:

await fetch('https://api.indexnow.org/indexnow', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    host: 'yourdomain.com',
    key: 'your-indexnow-key',
    urlList: ['https://yourdomain.com/updated-page']
  })
});

Setup Tip: Generate your IndexNow key at bing.com/indexnow and host it at yourdomain.com/{key}.txt.

The JavaScript Rendering Problem

One more common eligibility failure: JavaScript-only content.

AI crawlers generally cannot execute intensive client-side JavaScript. If your content is rendered client-side — a React SPA, a Vue app, or content injected via useEffect — the crawler sees an empty shell.

The fix is server-side rendering (SSR) or static generation for all content you want AI-indexed. For Cloudflare Workers deployments:

// Return pre-rendered HTML, not a JS bundle shell
export default {
  async fetch(request) {
    return new Response(renderToString(<App />), {
      headers: { 'Content-Type': 'text/html' }
    });
  }
}

If full SSR is not practical, ensure at minimum that your <head> metadata, primary headings, and opening content paragraphs exist in the static HTML source — not injected by JavaScript after load.

The Eligibility Checklist

Before spending time optimizing your content for AI citation, verify your backend configurations match this checklist:

Requirement	ChatGPT Search	Google AI Overviews	Perplexity
Submitted to Bing Webmaster Tools	✅ Required

| — | — |
| Submitted to Google Search Console | — | ✅ Required

| — |
| OAI-SearchBot Not Blocked | ✅ Required

| — | — |
| Googlebot / Google-Extended Not Blocked | — | ✅ Required

| — |
| PerplexityBot Not Blocked | — | — | ✅ Required

|
| Cloudflare "Block AI Bots" Exceptions Set | ✅ Required

| Check

|
| robots.txt Wildcards Clear | ✅ Required

| ✅ Required

|
| Content in Static HTML Source | ✅ Required

| ✅ Required

|
| IndexNow / Crawler Hints Active | 📈 Recommended

| — | — |

After Eligibility: The Visibility Layer

Once your site clears eligibility, the visibility problem begins. This is what we are working on right now — applying this exact framework to Miami-Dade health practices, where 88% of health searches trigger a Google AI Overview (BrightEdge, 2026) and most independent practices are completely absent from the results.

If you are curious to know what we are working on right now, take a look at aeogeoai.net/local-ai-feature-miami

The primary driver of AI citation presence is not technical eligibility; it is brand mention volume — the number of independent, third-party indexed sources that reference the entity.

$$\text{AI Recommendation Probability} = \text{Citation Coverage} \times \text{Category Clarity} \times \text{Review Presence} \times \text{Third-Party Authority} \times \text{Evidence Consistency}$$

Brand mentions correlate 3x more strongly with AI citation than traditional backlinks: $0.664$ vs $0.218$ correlation coefficient. A site that is technically eligible but has zero third-party coverage across the web will clear the gate and remain completely invisible.

About the Author

I build open diagnostic utilities focused on semantic validation and AI search visibility. If you want to check how your brand, platform variables, or local entity mappings appear across ChatGPT, Claude, and Gemini simultaneously — testing both eligibility and visibility parameters — you can run a free, no-account check at aeogeoai.net.

The tool returns $0\text{–}100$ scores per model and supplies word-for-word response excerpts showing exactly what each system says about your platform. Three free checks per day, no signup required.