Why Your Brand Has Good SEO But Zero AI Citations

Good SEO does not translate to AI citations because Google and AI answer engines run on different signals. AI models cite brands based on crawler access, entity resolution, content structure, third-party consensus, and search-index coverage; none of which correlate reliably with Google rankings. Each cause is diagnosable and independently fixable.

It came in at 4pm on a Tuesday, forwarded by the head of sales with no subject line. A prospect had typed something like "best tools for AI brand visibility" into ChatGPT and screenshotted the response. Three competitors, each with a tidy two-sentence description. Your company nowhere in it. The sales lead's message was four words: "Can we fix this?"

You open your SEO dashboard to look for answers and find none. Page one on the terms that matter. Eighty-plus blog posts. A Core Web Vitals score that most agencies would be proud of. By every metric you track, you are doing the work. The work is not the problem.

The gap between good SEO and AI citation is technical, not qualitative. The five causes below are the ones BrandViz.AI surfaces most often when auditing new clients. Each is independent: a brand can be blocked on cause one and perfectly fine on causes two through five. Fixing only the ones that actually apply is faster and more effective than a broad content push.

Cause 1: AI crawlers cannot access your site

Before any AI model can learn from your content, a crawler has to read it. The major AI platforms each run their own: OpenAI uses GPTBot for training data andOAI-SearchBot for ChatGPT Search retrieval. Anthropic uses ClaudeBot. Google uses Google-Extended for Gemini. Perplexity uses PerplexityBot. If any of these are blocked, the corresponding platform cannot cite your content, regardless of how well-written it is.

Three places where blocks appear without anyone realising:

• robots.txt: an overly broad Disallow: / rule, a wildcard block on unknown bots, or an explicit deny on User-agent: GPTBot added during an early "block AI scrapers" trend that nobody removed.
• WAF rules: web application firewalls from providers like Cloudflare can classify AI crawlers as suspicious traffic. Cloudflare's Bot Fight Mode, in particular, blocks a wide range of automated agents by default. It is common for a site to have Bot Fight Mode enabled as a security setting without anyone connecting it to AI visibility.
• Geo-blocking: CDN or hosting rules that restrict access by geography can block AI crawlers whose infrastructure sits in a region your rules exclude, typically US datacenters for most of the major crawlers.

How to check: fetch your robots.txt file directly (yourdomain.com/robots.txt) and scan every Disallow rule. Look for rules that apply to * (all bots) and check whether GPTBot, OAI-SearchBot, ClaudeBot, and PerplexityBot are listed anywhere. If you are running Cloudflare, review your Bot Fight Mode and Firewall Rules settings. In BrandViz.AI audits, OAI-SearchBot blocks show up as the most common undiscovered issue among B2B SaaS companies who otherwise have clean technical SEO.

The fix: explicitly allow the crawlers you want. Add them to your robots.txt allowlist and review any WAF rules that target bot traffic.

Cause 2: Your brand is not a resolved entity

AI models think in entities, not keyword frequency. An entity is a named thing with consistent attributes: a brand has a name, a category, a set of capabilities, and a location. When a model encounters your brand name across multiple sources, it tries to anchor that name to a known entity in its understanding. If the signals are inconsistent, the model cannot form a confident picture and tends to skip you in favour of brands where the picture is clear.

Inconsistency shows up in predictable places. Your homepage calls the product an "AI visibility platform." Your G2 profile says "LLM monitoring tool." Your LinkedIn company page says "B2B SaaS analytics." Your schema markup has no description field at all. Each source is sending a different signal about what your brand is, and the model is left to guess which one to trust. In practice, it defaults to the competitors whose signals are consistent.

The fix requires three things working together: an Organisation schema block on your homepage that explicitly states your name, category, description, and founding location; sameAs links pointing to your LinkedIn, G2, Crunchbase, and Wikipedia pages; and consistent terminology across every external profile so all sources describe the same entity in the same terms. For a deeper walkthrough on entity resolution, our GEO guide covers the technical schema requirements in full.

Cause 3: Your content describes features, not answers questions

ChatGPT is not a search engine returning pages ranked by relevance. It is composing an answer, and it pulls from whatever content most directly answers the question it was asked. Content built around product features and benefit statements rarely survives that process: it does not map to a specific question, so the model passes over it. Content that opens each section with a direct answer to a buyer question gets pulled in cleanly, because the model can see exactly what it needs.

The difference is structural, not qualitative. Two pages can cover identical ground, and the one that opens each section with a direct answer to a buyer question will consistently outperform the one that opens with a feature description. A page that opens with "Our platform provides real-time AI visibility monitoring across ChatGPT, Claude, Gemini, and Perplexity" is describing a feature. A page that opens with "To find out whether ChatGPT is recommending your competitors, you need to simulate the specific buying queries your prospects use across each AI platform and trace which sources each model cites" is answering a question.

The practical audit: pick ten of your most important pages and read the first paragraph of each one. If it describes what your product does rather than answering a specific question a buyer might ask, restructure it. For step-by-step guidance on the structural changes that move the needle most, our content writing guide walks through the six decisions that determine whether AI can cite a page.

Cause 4: Third-party sources do not agree on what you do

Before a model recommends a brand, it has read about that brand from multiple directions: your own site, G2, Capterra, LinkedIn, comparison articles, industry newsletters. If those sources all say roughly the same thing, the model cites with confidence. If they conflict, or if some are missing entirely, the model treats you as lower-confidence and defaults to whoever has the cleaner signal. You get skipped not because you are unknown, but because the sources that know you cannot agree.

This is distinct from entity resolution (Cause 2), which is about consistency within your own channels. Third-party consensus is about what happens when the model reads sources you do not control. A brand can have perfect schema markup and a clean G2 profile, then get undercut because a widely-cited comparison article in an industry publication misclassifies the product, or because the brand is absent from the category listings those sources rely on.

The fix: audit the top five third-party sources that cover your category. Check that your brand name, category label, and core capability description match across all of them. Where you are missing entirely, prioritise getting listed. Where descriptions are outdated or wrong, most platform editors will accept correction requests. The goal is that any model reading those sources arrives at the same one-sentence understanding of what your brand does.

Cause 5: ChatGPT retrieves from Bing, not Google

This is the most specific and least understood cause, and the one that surprises teams with strong Google performance the most. ChatGPT Search, the live-retrieval layer that powers real-time responses in ChatGPT, uses Bing's search index via OAI-SearchBot. OpenAI documents this directly: when ChatGPT searches the web to form a response, it is querying Bing, not Google.

The implication is direct: a brand that has excellent Google rankings but has never submitted to Bing Webmaster Tools, or has blocked OAI-SearchBot in its robots.txt, can be completely invisible to ChatGPT's retrieval layer. Every optimisation made for Google's crawler has no bearing on whether Bing has indexed your pages or whether OAI-SearchBot can access them.

Bing Webmaster Tools is a separate submission process from Google Search Console. Many B2B SaaS teams submitted to Google when they launched and have never touched Bing. If that describes your situation, this cause alone may account for a large fraction of your ChatGPT invisibility.

The fix, in order:

1. Confirm OAI-SearchBot is not blocked in your robots.txt (see Cause 1).
2. Submit your site to Bing Webmaster Tools and submit your sitemap there directly.
3. Implement IndexNow, the push-based indexing protocol supported by Bing, Yandex, Naver, and Seznam. IndexNow lets you notify Bing's index immediately when you publish or update a page, rather than waiting for the crawler to find it. Over 80 million websites use it. Most CMS platforms have a plugin or native integration.

This is the most common hidden blocker we see in BrandViz.AI audits for companies with otherwise healthy SEO. Google visibility and Bing visibility are independent systems. ChatGPT depends on one of them, and it is not the one most teams have invested in.

The 5-minute self-audit: check which causes apply to you

Not all five causes will apply to your site. Run through these questions against your own setup before deciding where to spend time.

If you answered "no" to more than two of these, you have concrete, specific work to do before content volume or quality makes a material difference to your AI citations. Fix the infrastructure first.

What SEO optimises for vs what AI citation requires

The underlying reason these five causes exist is that Google and AI answer engines use fundamentally different signals. Optimising for one does not carry over to the other.

The broader context on why these two channels diverged, and what a strategy that addresses both looks like, is in our SEO vs GEO analysis. If you arrived at this post from a "why are my competitors in ChatGPT" search, the companion piece on competitor visibility covers how to diagnose which specific sources are driving those recommendations.

Frequently Asked Questions

Does blocking GPTBot affect my Google rankings?

No. GPTBot is OpenAI's training data crawler and has no relationship with Googlebot. Blocking GPTBot prevents OpenAI from including your content in future training runs, but it has no effect on your Google rankings. Blocking OAI-SearchBot, however, prevents ChatGPT Search from retrieving your pages in real-time responses, which directly affects your ChatGPT citation rate.

Does fixing these five causes guarantee AI citations?

Fixing these causes removes the blockers that prevent AI models from seeing and citing your brand. It does not guarantee citations in any specific response: AI models compose answers contextually, and even a well-optimised brand will not appear in every relevant query. The realistic goal is consistent presence in the queries that matter most for your category, which requires tracking those queries systematically across models.

Does a high Google ranking help at all with AI visibility?

For Gemini, indirectly: Gemini draws on Google's index, so pages that rank well on Google are more likely to be in Gemini's retrieval pool. For ChatGPT, Perplexity, and Claude, your Google ranking has little bearing on whether you appear in responses. Each platform has its own retrieval and training pipeline. Bing Webmaster Tools submission matters more for ChatGPT than any Google-side optimisation.

How often do robots.txt blocks on AI crawlers show up in audits?

In BrandViz.AI's experience auditing B2B SaaS sites, OAI-SearchBot blocks are the single most common undiscovered technical issue among companies with otherwise strong SEO. They typically originate from either a blanket wildcard block added during the "block AI scrapers" trend of 2023, or from Cloudflare Bot Fight Mode enabled as a security setting without AI visibility implications being considered.

What is entity resolution and why does it affect AI citations?

Entity resolution is the process by which AI models connect your brand name to a consistent set of attributes: what you do, what category you are in, where you are based, and what makes you distinct. When your schema markup, G2 profile, LinkedIn page, and homepage all use the same terminology, models can build a confident picture and cite you with attribution. When the signals conflict, models default to brands where the picture is clear. Schema markup with a populated Organisation block and sameAs links is the most direct way to establish this.

Should I submit to Bing Webmaster Tools even if my audience is primarily in regions where Bing is less popular?

Yes. Bing's market share in your users' geography is irrelevant to this question: ChatGPT Search uses Bing's index regardless of where the query originates. A buyer in Australia using ChatGPT triggers OAI-SearchBot against Bing's index, not against Google's. Bing Webmaster Tools submission is a ChatGPT visibility requirement, not a Bing search engine strategy.