In August 2026, Google's AI search told users that a Japanese company was bankrupt. It was not. The false claim spread fast enough to spark investor panic before anyone could correct it.
This is not a one-off glitch. AI search engines are inventing facts about businesses every day. Wrong addresses. Wrong hours. Wrong ownership. Wrong products. Claims that competitors closed down. Recommendations that mix up two businesses with similar names. These errors appear in ChatGPT answers, Google AI Overviews, and Perplexity responses, and the businesses being misrepresented usually have no idea it is happening.
The IAB addressed this directly in its August 3, 2026 framework. It called the dimension Portrayal. Here is what Portrayal means, why it matters more than most brands realize, and what you can actually do about it.
What the IAB calls Portrayal
The IAB framework organizes AI visibility into four dimensions: Presence, Prominence, Portrayal, and Persuasion. Portrayal is the one most brands have never heard of, and it is the one that can hurt you the most.
Presence asks whether your brand appears in AI answers at all. Prominence asks how high you rank. Persuasion asks whether any of it drives revenue. Portrayal asks a different question entirely: when an AI engine mentions your business, is what it says accurate?
The IAB framework defines Portrayal through two metrics. Hallucination Rate measures how often an AI engine states a fact about a brand that is demonstrably false. Factual Inaccuracy Rate measures how often the details are wrong even when the overall mention is correct. A wrong phone number counts. A wrong location counts. A claim that you were acquired when you were not counts.
Traditional SEO never had to worry about this. Google showed your title tag and meta description, and users clicked through to read your actual page. The search engine displayed what you published. AI search engines do not display your page. They synthesize an answer from multiple sources, fill in gaps with their own reasoning, and sometimes fabricate details that none of those sources actually said. Your brand is being described, and you did not write the description.
Real examples from this month
The Japanese bankruptcy case is dramatic, but it is not isolated. A study of high street retailers in Leeds, England found that more than half had inaccurate information appearing in AI chatbot responses about their businesses. Wrong details about products, services, locations, and operating status. These are small businesses that depend on foot traffic and local search, and AI engines were actively misinforming their potential customers.
In July 2026, Coinbase experienced an AI hallucination that sent football fans into a frenzy when AI-generated content provided incorrect information about a promotion. The UK Home Office was caught using AI-hallucinated information to deny an asylum claim, a case serious enough that a judge flagged it directly. These incidents are happening across industries, and the pace is accelerating.
Profound, a competitor in the AI visibility space, launched a product called FactCheck in July 2026 specifically to measure the accuracy of AI answers about brands. The fact that a venture-backed company built an entire product around this problem tells you the hallucination issue is not theoretical. It is a market.
Why AI engines hallucinate about your business
AI hallucination is not random. It happens for specific, identifiable reasons. Understanding these causes is the first step toward fixing them.
First, inconsistent entity data. Your business name appears differently across the web. Google Business Profile has one version. Yelp has another. Your website has a third. The Better Business Bureau has a fourth. When AI engines try to build a unified picture of your business from these conflicting sources, they resolve the contradictions by guessing. Sometimes they guess wrong.
Second, missing or incomplete schema markup. Structured data tells search engines exactly what your business is, what it sells, where it is located, and how to contact you. Without it, AI engines have to infer these details from unstructured page text. Inference is where errors creep in. If your Organization schema is missing or your LocalBusiness schema has outdated hours, the AI fills in what it thinks is right.
Third, crawl blocks. If your robots.txt blocks AI crawlers, the engines cannot read your site directly. They rely on third-party data aggregators, directory sites, and cached versions of your content. Those sources may be outdated, incomplete, or simply wrong. The AI cannot verify against the primary source because you blocked it from reading the primary source.
Fourth, thin or ambiguous content. If your website has limited information about what you do, where you are, and what makes you different from a similarly named business, the AI has to extrapolate. Extrapolation from thin data produces confident-sounding errors.
The cost of wrong information
Portrayal errors are not just embarrassing. They cost money. If an AI engine tells a potential customer that your restaurant is closed on Saturdays when you are open, that customer goes somewhere else. You never see the lost sale. You do not even know it happened.
If the AI confuses your business with a competitor that has a similar name, your traffic goes to them. If it lists services you do not offer, you get unqualified inquiries that waste your time. If it states you were acquired or closed, the damage can affect investor confidence, supplier relationships, and hiring. The Japanese company that Google AI falsely called bankrupt experienced all three of these in a single news cycle.
The IAB framework recognizes this cost by placing Portrayal alongside Presence and Persuasion as a measurement pillar. A brand can have excellent Presence and terrible Portrayal. Being mentioned frequently is not helpful if the mentions are wrong.
How to check what AI engines are saying about you
Start with the direct approach. Open ChatGPT, Perplexity, and Google AI Overviews. Ask questions a potential customer would ask: What does this business do? Where is it located? What are its hours? Who owns it? Is it still open? Compare every answer to reality. Note every error.
This manual check gives you a snapshot, but it is not systematic. AI responses change between sessions. The same query can return different answers depending on the model version, the user's location, and the conversation context. To track Portrayal over time, you need a repeatable process.
That is what Parceit checks. Our audit examines the signals that cause hallucinations before they happen. We look at whether your robots.txt allows AI crawlers to read your site directly. We check whether your Organization and LocalBusiness schema exist and whether they match the information on your actual pages. We verify that your entity data is consistent across the sources AI engines use to build their picture of your business.
Every check is visible and reproducible. The same site produces the same result every time. That is what the IAB calls decision-grade measurement, and it is the tier you need when Portrayal errors are costing you customers you cannot see.
What to fix first
If you find Portrayal errors, the fixes follow a clear order.
- Fix your entity data. Make sure your business name, address, phone number, hours, and category are identical across Google Business Profile, your website, and every major directory. Consistency is the single highest-impact Portrayal fix.
- Add or correct your Organization and LocalBusiness schema. These structured data types tell AI engines exactly what your business is. Without them, engines guess.
- Remove crawl blocks for AI engines. Check your robots.txt for GPTBot, PerplexityBot, Google-Extended, and ClaudeBot. If they are blocked, the AI cannot read your site and has to rely on third-party sources that may be wrong.
- Expand thin content. If your website has limited detail about what you do, add clear, factual descriptions. The more primary information AI engines can read directly, the less they need to infer.
- Monitor regularly. Portrayal is not a one-time fix. AI models update, new sources appear, and entity data drifts. Checking once is better than never checking, but ongoing monitoring is what catches new errors before they spread.
The brand safety layer you did not know you needed
Traditional SEO was about controlling what appeared on the search results page. You optimized your title tags, wrote your meta descriptions, and managed your Google Business Profile. The search engine displayed what you gave it.
AI search changed the model entirely. Now the engine writes its own description of your business, and you do not get to review it before it reaches a potential customer. That description may be accurate. It may also be catastrophically wrong, and you will not know unless you check.
Portrayal is the brand safety layer traditional SEO never needed. It is the difference between being listed correctly and being the subject of a confident, detailed, and entirely fabricated claim that reaches thousands of potential customers before you catch it.
Run a free audit at parceit.com. We check the crawlability, entity data, and schema signals that determine whether AI engines can build an accurate picture of your business. Every signal is visible, every result is reproducible, and every error has a clear path to fix.