78% of B2B companies appear in no responses generated by the main AI engines. Behind this figure lies a more nuanced reality: most of these companies aren't invisible because they lack expertise or quality. They're invisible because they make specific mistakes — often technical, often recent, often unintentional.

This article lists the 10 most common mistakes observed in the market in 2025-2026, with a diagnosis and concrete fix for each.

Mistake 1 — Treating all AI engines as a single channel

This is the most widespread and strategically costly mistake. An analysis of 680 million citations revealed that only 11% of domains are cited by both ChatGPT and Perplexity. The two platforms draw from almost entirely different sources.

Concretely: a company that optimises for ChatGPT can be perfectly invisible on Perplexity — and vice versa. Each AI engine has its own citation logic, its own retrieval engine, and its own authority signals.

  • ChatGPT → relies on Bing for web search, favours well-documented entities
  • Perplexity → favours earned media and content freshness
  • Gemini → anchored in Google Search, values E-E-A-T signals
  • Copilot → exclusively on the Bing index, values LinkedIn and the Microsoft ecosystem
  • Claude → uses Brave Search, requires source verifiability

The fix: build a multi-platform AI visibility strategy, with specific optimisations for each engine — not a one-size-fits-all approach.

Mistake 2 — Blocking AI crawlers in robots.txt

Since 2024, many companies have added rules to their robots.txt to block AI crawlers and protect their training data. This is a legitimate decision — but often poorly executed.

The problem: there is a critical difference between training crawlers (which collect data to train models) and search crawlers (which determine what the model cites in real time). Blocking the wrong crawler can make you invisible without realising it.

Examples of crawlers to distinguish:

  • Anthropic: ClaudeBot (training) vs Claude-SearchBot (citations) — blocking the first is fine, blocking the second = invisible on Claude
  • OpenAI: GPTBot (training) vs OAI-SearchBot (citations) — same logic
  • PerplexityBot: a single crawler, blocking it = invisible on Perplexity

The fix: check your robots.txt and explicitly authorise the search crawlers of each platform.

Mistake 3 — Confusing Google ranking with AI visibility

This is the most persistent myth. "I'm on the first page of Google, so I must appear in AI responses." The data says otherwise.

In mid-2025, 76% of Google AI Overview citations came from the organic top 10. By early 2026, this figure had fallen to 38% according to Ahrefs, and as low as 17% according to BrightEdge. The correlation between Google ranking and AI visibility is rapidly eroding.

On ChatGPT and Perplexity, the situation is even clearer: a 2026 study of 150 B2B SaaS companies found that 44% of brands in the Google top 10 receive zero ChatGPT citations for the same keywords.

The fix: measure AI visibility separately from SEO, with manual tests on each platform or dedicated tracking tools.

Mistake 4 — Publishing content "for humans" without thinking about AI extraction

AI engines don't read your content like a human. They "chunk" — they break your text into blocks and extract passages that directly answer a question. Content written in long narrative paragraphs without clear structure provides few extractible passages.

What LLMs can extract easily:

  • A direct answer in the first 2 sentences under an H2 heading
  • A comparison table in clean HTML
  • A statistic with its source and date in the same sentence
  • A structured list with standalone elements

What they cannot extract:

  • An argument developed over 4 paragraphs without a clear conclusion
  • A value buried in a long introduction
  • Content rendered in JavaScript that only displays on load

The fix: reformat key pages with the "direct answer first, development second" principle — under each H2 heading, the answer in maximum 2 sentences before any context.

Mistake 5 — Neglecting Schema.org structured data

A Yext analysis of 6.8 million AI citations in 2025 reveals that 42% of citations come from directory listings and structured data — not narrative web pages. Schema.org markup is a layer that LLMs can read directly, independently of text.

The most impactful schemas for AI visibility:

  • Organization on the homepage — tells LLMs who you are
  • Product or Service on offer pages — tells what you offer and at what price
  • FAQPage on FAQ pages — cited at significantly higher rates on question-form queries
  • Article with dateModified on blog articles — direct freshness signal

The fix: implement at minimum Organization + FAQPage + Article with dateModified on your site.

Mistake 6 — Ignoring Bing and the Brave index

Most French-speaking companies have optimised everything for Google and never done anything for Bing or Brave. Yet:

  • 87% of ChatGPT Search citations correspond to top Bing results
  • 86.7% of Claude citations correspond to top Brave Search results
  • Copilot works exclusively on the Bing index

Not being indexed on Bing or Brave means being invisible on ChatGPT Search, Claude and Copilot simultaneously — three of the five main AI engines.

The fix: create a Bing Webmaster Tools account, submit your sitemap, activate IndexNow. Create a Brave Webmaster Tools account. These two actions take less than an hour.

Mistake 7 — Having content without author signals

AI engines — particularly Claude and Gemini — evaluate the credibility of the sender as much as the quality of the content. An article without an identifiable author, without a bio, without a link to a verifiable professional profile is systematically disadvantaged.

Data observed on Claude shows that an author with 10+ years of declared experience in the field increases the probability of citation by 45%. For Gemini, author signals are directly linked to the E-E-A-T criteria that Google has formalised.

The fix: add an identifiable author with a detailed bio and a LinkedIn link on every article. Implement the JSON-LD Person markup for each author.

Mistake 8 — Only publishing on your own site

LLMs trust sources that are themselves cited elsewhere. A company that only publishes on its own blog — without mentions in third-party sources — cannot be independently verified by AI engines.

The impact is documented and massive on Perplexity: 47% of its citations come from journalistic sources, and 89%+ are earned media (Fullintel-UConn, February 2026). On Gemini, press articles republished on multiple third-party sites see their citations increase by up to 325% (Stacker/Scrunch).

The fix: identify the 5 publications that LLMs regularly cite in your sector and build a strategy to be mentioned there — expertise contributions, citable studies, interviews.

Mistake 9 — Letting your content age without updates

AI engines favour freshness far more aggressively than Google. Content cited by LLMs is on average 25.7% more recent than traditional organic results (Ahrefs, 2025). Perplexity cites content published in the last 30 days at an 82% rate (2026 analysis).

An excellent article published 18 months ago without an update is progressively excluded from citation candidates — even if its content remains relevant.

The fix: set up a content update schedule. Add a visible "Last updated" date, update key statistics, and add the dateModified property in the JSON-LD of each article.

Mistake 10 — Not measuring AI visibility

This is the mistake that perpetuates all the others. Only 16% of brands systematically track their performance on AI platforms (Erlin, 2026). The remaining 84% don't know if they're visible or not — and therefore can't fix what isn't working.

Without measurement, it's impossible to know:

  • On which engines you appear and on which you're absent
  • Which competitors are getting citations in your place
  • Whether your optimisations have a real impact
  • How you're described when you're cited

The fix: manually test your visibility on the 5 main AI engines at least once a month, with a fixed set of prompts representing your prospects' queries. Our free scoring tool automates this measurement across all 5 engines simultaneously.

The 3 most urgent mistakes to fix

If you need to prioritise, here's the order of immediate impact:

1. Check robots.txt (5 minutes, impact in 2-4 weeks) — the most common self-inflicted wound.

2. Register on Bing Webmaster Tools and Brave Webmaster (1 hour, impact within weeks) — unlock visibility on 3 AI platforms simultaneously.

3. Add Schema.org Organization + FAQPage (a few hours, impact in 2-4 weeks) — the most direct structural signal for all LLMs.

The other mistakes require more work but have longer impact timelines — start with these three.

Where to start?

Our free scoring tool lets you measure your current visibility on the 5 main AI engines in a few minutes — and identify the most critical mistakes for your specific situation.

For a complete audit with a prioritised correction plan, our AI Diagnostic identifies precisely which of these 10 mistakes affect you and in what order to fix them.

To go deeper on each engine, see our complete series: Why ChatGPT ignores your business, Why Perplexity doesn't cite your website, Why Gemini doesn't recommend your business, Why Copilot doesn't mention your business, Why Claude doesn't talk about you.

Sources: analysis of 680 million AI citations (2025-2026), Ahrefs study on Google ranking / AI Overview citation correlation (2025-2026), Fullintel-UConn study on Perplexity citation patterns (IPRRC, February 2026), Erlin data on brand AI performance tracking (2026), Yext analysis of 6.8 million AI citations (October 2025).