{"id":8029,"date":"2026-09-29T07:30:00","date_gmt":"2026-09-29T05:30:00","guid":{"rendered":"https:\/\/taismo.de\/?p=8029"},"modified":"2026-09-24T14:49:45","modified_gmt":"2026-09-24T12:49:45","slug":"ai-hallucination","status":"publish","type":"post","link":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/","title":{"rendered":"AI hallucination"},"content":{"rendered":"<style>\n.taismo-art{display:grid;grid-template-columns:320px minmax(0,1fr);gap:36px;align-items:start}\n.taismo-art .toc-sticky{position:sticky;top:80px}\n.taismo-art .toc-sticky ol li{margin-bottom:9px}\n.taismo-art .toc-sticky ol a:hover{color:var(--toc-hover) !important}\n.taismo-art .taismo-main h2{scroll-margin-top:90px;margin-top:38px}\n.taismo-art .lead-def{font-size:1.15em;line-height:1.6;background:#f7f2ec;border-left:4px solid #ED8924;padding:16px 20px;border-radius:8px}\n.taismo-art table{border-collapse:collapse;width:100%;margin:18px 0}\n.taismo-art th,.taismo-art td{border:1px solid #d9e0df;padding:9px 12px;text-align:left;font-size:0.95em}\n.taismo-art th{background:#072328;color:#fff}\n.taismo-art figure{margin:26px 0}\n.taismo-art figcaption{font-size:0.9em;color:#55605f;margin-top:8px}\n.taismo-art pre{background:#072328;color:#f5f5f5;padding:16px 18px;border-radius:10px;overflow-x:auto;font-size:0.86em;line-height:1.5}\n@media(max-width:900px){.taismo-art{grid-template-columns:1fr}.taismo-art .toc-sticky{position:static}}\n<\/style>\n<div class=\"taismo-art\">\n<aside class=\"toc-sticky\" style=\"--toc-hover:#ED8924;background:#072328;border-radius:14px;padding:22px 24px\">\n<div style=\"text-align:center;margin-bottom:14px\"><svg fill=\"#ffffff\" role=\"img\" aria-label=\"taismo Wiki\" style=\"width:200px;max-width:80%;height:auto;display:inline-block\" id=\"Ebene_1\" data-name=\"Ebene 1\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 209.04 52.32\"> <g id=\"Layer_2\" data-name=\"Layer 2\"> <g id=\"wikipedia\"> <path d=\"M97.08,27.77h-17.76c-.77,0-1.4.63-1.4,1.4v17.76c0,.77.63,1.4,1.4,1.4h17.76c.77,0,1.4-.63,1.4-1.4v-17.76c0-.77-.63-1.4-1.4-1.4ZM97.22,33.22c0,.25-.21.46-.46.46h-.67c-.07.03-.13.07-.19.11-.19.18-.33.41-.4.66,0,0,0,.01,0,.02-.03.07-2.9,7.29-3.87,9.67v.02c-.09.17-.19.33-.31.48-.15.2-.38.32-.63.33-.56,0-.84-.63-.93-.84-.28-.62-.79-1.83-1.51-3.57-.55,1.24-1.25,2.8-1.64,3.59h0c-.19.36-.49.83-.97.83-.56,0-.82-.63-.9-.84h0c-.49-1.25-1.23-2.99-1.95-4.68-.79-1.83-2.39-5.73-2.43-5.78h-.72c-.25,0-.46-.21-.46-.46v-.58c0-.25.21-.45.46-.46h4.13c.25,0,.46.21.46.46v.58c0,.25-.21.45-.46.46h-.82s.03.09.05.14c.59,1.61,2.36,5.75,3.1,7.47l1.28-2.81c-.15-.36-2.12-4.68-2.2-4.8h-.4c-.25,0-.46-.21-.46-.46v-.58c0-.25.21-.45.46-.46h2.98c.25,0,.46.21.46.46v.58c0,.25-.21.46-.46.46h-.1c.06.14.94,2.12.94,2.12,0,0,.93-2.06.95-2.12h-.39c-.25,0-.46-.21-.46-.46v-.58c0-.25.21-.45.46-.46h2.81c.25,0,.46.21.46.46v.58c0,.25-.21.45-.46.46h-.43s-2.06,4.08-2.06,4.08l1.56,3.51s2.9-7.59,2.9-7.6h-.96c-.25,0-.46-.21-.46-.46v-.58c0-.25.21-.45.46-.46h3.78c.25,0,.46.2.46.46v.58Z\"\/> <\/g> <\/g> <text font-family=\"Arial,Helvetica,sans-serif\" font-size=\"27.79\" font-weight=\"800\" transform=\"translate(99.44 47.64) scale(.99 1)\"><tspan x=\"0\" y=\"0\">IKI<\/tspan><\/text> <g id=\"Gruppe_796\" data-name=\"Gruppe 796\"> <path id=\"Pfad_1293\" data-name=\"Pfad 1293\" d=\"M6.92,16.77h-3.56v-6.56h3.66V3.37h8.04v6.84h5.69v6.56h-5.58v8.97c-.08,1.05.13,2.1.6,3.04.4.6,1.24.9,2.51.9.54,0,1.01,0,1.39-.03.41-.02.82-.07,1.23-.13v6.35c-.75.19-1.52.31-2.3.36-1.02.09-1.95.13-2.79.13-1.41.04-2.81-.17-4.15-.6-1.07-.36-2.02-1.01-2.76-1.86-.76-.95-1.29-2.06-1.54-3.25-.34-1.55-.49-3.14-.46-4.73v-9.14h0Z\"\/> <path id=\"Pfad_1294\" data-name=\"Pfad 1294\" d=\"M44.43,35.91h-7.06l-.44-1.97c-.68.72-1.48,1.34-2.35,1.8-.95.46-2.01.69-3.07.66-.99,0-1.98-.1-2.95-.33-.92-.21-1.78-.63-2.51-1.23-.77-.66-1.37-1.49-1.75-2.43-.48-1.28-.71-2.63-.66-3.99v-1.42c-.04-1.17.2-2.33.71-3.39.44-.87,1.09-1.63,1.89-2.19.82-.56,1.74-.96,2.71-1.18,1.06-.24,2.14-.36,3.23-.36.87,0,1.72.06,2.57.22.72.14,1.41.36,2.08.66v-.93c0-.54-.07-1.07-.22-1.59-.16-.48-.46-.91-.87-1.2-.55-.38-1.18-.63-1.83-.77-1-.2-2.01-.29-3.04-.28-.91,0-1.81.09-2.71.25-.97.16-1.8.32-2.49.46v-6.12c.48-.13.96-.22,1.45-.28.57-.07,1.16-.14,1.78-.22.62-.07,1.22-.13,1.78-.16.57-.04,1.04-.05,1.45-.05,2.1-.05,4.18.19,6.21.71,1.4.37,2.69,1.08,3.74,2.08.92.94,1.57,2.13,1.86,3.42.34,1.54.51,3.1.49,4.68v15.15h.01,0ZM33.71,30.83c1.63.12,3.04-1.1,3.16-2.72v-2.58c-.77-.58-1.72-.87-2.68-.82-.85-.04-1.69.16-2.43.57-.6.38-.9,1.16-.9,2.33v.38c-.07.78.16,1.55.66,2.16.6.51,1.39.76,2.19.69\"\/> <rect id=\"Rechteck_760\" data-name=\"Rechteck 760\" x=\"49.46\" y=\"10.21\" width=\"8.26\" height=\"25.71\"\/> <path d=\"M131.36,22.82v-.42c-.06-1.6.12-3.21.52-4.76.35-1.1,1.23-1.64,2.65-1.64.59-.03,1.16.11,1.66.41.43.29.75.73.9,1.23.2.66.32,1.34.36,2.02.05.81.08,1.72.08,2.74v.42h8.38v-1.01c.04-1.92-.23-3.83-.82-5.66-.48-1.43-1.27-2.73-2.33-3.8-1.02-.99-2.27-1.73-3.64-2.13-1.53-.45-3.12-.67-4.71-.66-1.57-.01-3.12.2-4.62.66-1.36.41-2.6,1.14-3.61,2.13-1.06,1.07-1.86,2.37-2.35,3.8-.61,1.82-.89,3.74-.85,5.66v1.01h8.38Z\"\/> <path d=\"M93,22.82v-2.38c-.06-.88.16-1.76.63-2.51.39-.57,1.04-.89,1.72-.87.26,0,.51.03.77.08.26.06.51.19.71.38.25.24.43.54.52.87.15.51.21,1.03.19,1.56v2.88h8.2v-2.7c-.05-.8.16-1.6.6-2.27.4-.52,1.03-.82,1.69-.79.54,0,1.07.17,1.51.49.45.33.69,1.13.69,2.41v2.86h8.2v-4.99c.03-1.32-.17-2.64-.6-3.89-.34-.96-.9-1.83-1.64-2.54-.69-.63-1.51-1.11-2.41-1.39-.92-.3-1.88-.44-2.84-.44-1.16-.02-2.3.21-3.36.69-1.09.55-2.07,1.3-2.87,2.22-.57-.95-1.38-1.72-2.35-2.24-1.16-.49-2.41-.72-3.66-.66-1.18-.04-2.35.22-3.39.76-.99.57-1.88,1.31-2.63,2.19l-.54-2.3h-7.34v12.61h8.2Z\"\/> <path d=\"M73.21,26.15c0-1.84,1.49-3.33,3.33-3.33h2.18c-.87-.85-1.94-1.47-3.12-1.8l-1.75-.6c-.73-.25-1.35-.46-1.86-.63-.43-.13-.85-.3-1.26-.49-.28-.13-.52-.33-.71-.57-.16-.28-.24-.6-.22-.93,0-.22.04-.45.13-.66.13-.25.34-.46.6-.57.43-.2.89-.34,1.36-.41.81-.13,1.62-.18,2.43-.16.92,0,1.85.07,2.76.22.82.11,1.63.28,2.43.49v-6.12c-.48-.13-.95-.22-1.45-.28-.56-.07-1.16-.14-1.78-.22-.61-.07-1.21-.13-1.78-.16s-1.05-.05-1.45-.05c-3.72,0-6.51.75-8.37,2.27-1.78,1.35-2.82,3.48-2.79,5.71-.03,1.07.11,2.13.41,3.15.24.76.64,1.46,1.18,2.05.52.56,1.15,1.01,1.83,1.34.79.38,1.6.72,2.43,1.01l1.26.44c.83.28,1.63.61,2.41,1.01.47.28.75.8.71,1.34.07.72-.35,1.4-1.04,1.64-1.23.31-2.51.45-3.77.38-.95,0-1.9-.09-2.84-.25-1.06-.16-1.93-.31-2.63-.46v6.12c.5.13,1,.22,1.51.28.6.07,1.23.14,1.89.22.66.07,1.3.13,1.92.16.62.03,1.13.05,1.53.05,1.5.03,3.01-.1,4.49-.36v-9.83Z\"\/> <\/g> <g id=\"e\"> <path d=\"M150.81,38.2c-.26-1.16-1.38-2.22-2.54-2.45-.71-.14-2.69-.32-3.99-.3-1.31-.01-3.28.16-3.99.3-1.16.23-2.29,1.29-2.54,2.45-.15.7-.29,1.66-.29,2.97,0,1.31.15,2.27.29,2.97.26,1.16,1.38,2.22,2.54,2.45.38.08,1.14.16,1.95.22.09.15.19.3.3.46.43.66.77,1.11,1.02,1.43.42.53,1.03.53,1.45,0,.25-.32.58-.78,1.02-1.43.11-.16.21-.32.3-.46.82-.06,1.57-.15,1.95-.22,1.16-.23,2.29-1.29,2.54-2.45.15-.7.29-1.66.29-2.97,0-1.31-.15-2.27-.29-2.97h0ZM145.03,45.13c-.03.16-.18.31-.34.34-.1.02-.23.03-.42.03-.18,0-.32-.02-.42-.03-.16-.03-.31-.18-.34-.34-.02-.1-.03-.23-.03-.42,0-.18.02-.32.03-.42.03-.16.18-.31.34-.34.1-.02.23-.03.42-.03.18,0,.32.02.42.03.16.03.31.18.34.34.02.1.03.23.03.42,0,.18-.02.32-.03.42ZM146.65,40.76c-.21.29-.51.56-.97.84-.58.36-.67.5-.67.79v.11c0,.41-.33.74-.74.74s-.74-.33-.74-.74v-.26c0-.34.08-.63.26-.88.17-.24.42-.46.8-.68.69-.41.82-.63.82-.94,0-.38-.32-.63-.83-.63-.55,0-.91.25-1.06.74-.1.32-.4.54-.75.54-.02,0-.04,0-.06,0-.24-.02-.46-.14-.6-.35-.14-.2-.17-.45-.1-.68.17-.52.49-.96.93-1.26.45-.3,1.01-.47,1.63-.47s1.24.21,1.69.58c.45.38.71.91.71,1.49,0,.4-.1.74-.32,1.04h0Z\"\/> <\/g> <g id=\"e-2\" data-name=\"e\"> <path d=\"M156.19,28.79c-.17-.75-.89-1.44-1.64-1.58-.46-.09-1.73-.2-2.58-.2-.84,0-2.12.1-2.58.2-.75.15-1.48.83-1.64,1.58-.09.45-.19,1.07-.19,1.92,0,.84.09,1.46.19,1.92.17.75.89,1.44,1.64,1.58.25.05.73.11,1.26.14.06.09.12.19.19.3.28.42.49.72.66.93.27.34.66.34.93,0,.16-.21.38-.5.66-.93.07-.11.13-.21.19-.3.53-.04,1.01-.09,1.26-.14.75-.15,1.48-.83,1.64-1.58.09-.45.19-1.07.19-1.92,0-.84-.09-1.46-.19-1.92h0ZM152.46,33.27c-.02.1-.11.2-.22.22-.06.01-.15.02-.27.02-.12,0-.2,0-.27-.02-.1-.02-.2-.11-.22-.22-.01-.06-.02-.15-.02-.27,0-.12,0-.2.02-.27.02-.1.11-.2.22-.22.06-.01.15-.02.27-.02.12,0,.2,0,.27.02.1.02.2.11.22.22.01.06.02.15.02.27,0,.12,0,.2-.02.27ZM153.51,30.45c-.13.19-.33.36-.63.54-.37.23-.43.32-.43.51v.07c0,.26-.21.48-.48.48s-.48-.21-.48-.48v-.17c0-.22.05-.4.17-.57.11-.16.27-.3.51-.44.45-.26.53-.41.53-.61,0-.25-.21-.4-.53-.4-.35,0-.59.16-.69.48-.07.21-.26.35-.48.35-.01,0-.02,0-.04,0-.16-.01-.3-.09-.39-.22-.09-.13-.11-.29-.06-.44.11-.34.32-.62.6-.81.29-.2.65-.3,1.05-.3s.8.13,1.09.37c.29.25.46.59.46.96,0,.26-.07.48-.21.67h0Z\"\/> <\/g> <g id=\"e-3\" data-name=\"e\"> <path d=\"M144.91,28.64c-.1-.44-.52-.84-.96-.92-.27-.05-1.01-.12-1.5-.11-.49,0-1.24.06-1.5.11-.44.08-.86.49-.96.92-.06.27-.11.63-.11,1.12,0,.49.05.85.11,1.12.1.44.52.84.96.92.14.03.43.06.74.08.03.05.07.11.11.17.16.25.29.42.38.54.16.2.39.2.54,0,.09-.12.22-.29.38-.54.04-.06.08-.12.11-.17.31-.02.59-.06.74-.08.44-.08.86-.49.96-.92.06-.27.11-.63.11-1.12,0-.49-.05-.85-.11-1.12h0ZM142.73,31.26c-.01.06-.07.12-.13.13-.04,0-.09.01-.16.01-.07,0-.12,0-.16-.01-.06-.01-.12-.07-.13-.13,0-.04-.01-.09-.01-.16,0-.07,0-.12.01-.16.01-.06.07-.12.13-.13.04,0,.09-.01.16-.01.07,0,.12,0,.16.01.06.01.12.07.13.13,0,.04.01.09.01.16,0,.07,0,.12-.01.16ZM143.34,29.61c-.08.11-.19.21-.37.32-.22.13-.25.19-.25.3v.04c0,.15-.12.28-.28.28s-.28-.12-.28-.28v-.1c0-.13.03-.24.1-.33.06-.09.16-.17.3-.26.26-.15.31-.24.31-.35,0-.14-.12-.24-.31-.24-.21,0-.34.09-.4.28-.04.12-.15.2-.28.2,0,0-.01,0-.02,0-.09,0-.17-.05-.23-.13-.05-.08-.06-.17-.04-.25.06-.2.18-.36.35-.47.17-.11.38-.18.61-.18s.47.08.63.22c.17.14.27.34.27.56,0,.15-.04.28-.12.39h0Z\"\/> <\/g> <\/svg><\/div>\n<h3 style=\"text-align:center;color:#fff;margin:0 0 14px;font-size:1.1em\">Contents<\/h3>\n<ol style=\"margin:0;padding-left:1.2em;line-height:1.5;font-weight:600;color:#fff;font-size:0.9em\">\n<li><a href=\"#term-profile\" style=\"color:#fff;text-decoration:none\">Term profile at a glance<\/a><\/li>\n<li><a href=\"#why-models-invent-facts\" style=\"color:#fff;text-decoration:none\">Why do language models invent facts?<\/a><\/li>\n<li><a href=\"#types\" style=\"color:#fff;text-decoration:none\">What types of AI hallucinations are there?<\/a><\/li>\n<li><a href=\"#semantic-drift\" style=\"color:#fff;text-decoration:none\">What is semantic drift and why does the plausible win?<\/a><\/li>\n<li><a href=\"#protect-your-brand\" style=\"color:#fff;text-decoration:none\">How do you protect your brand from false AI statements?<\/a><\/li>\n<li><a href=\"#faq\" style=\"color:#fff;text-decoration:none\">Frequently asked questions about AI hallucinations<\/a><\/li>\n<\/ol>\n<\/aside>\n<div class=\"taismo-main\">\n<h1 class=\"brand-claim\">What is an AI hallucination?<\/h1>\n<p class=\"lead-def\">An <strong>AI hallucination<\/strong> is a statement by an AI system that sounds convincing but is factually wrong or not backed by any source. The language model invents facts, figures, quotes, or sources and phrases them with the same confidence as correct information. AI hallucinations happen because a language model calculates the most likely continuation of a text and does not check whether its statement is true.<\/p>\n<figure>\n      <img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp\" alt=\"AI hallucination explained: why language models invent facts and prefer plausible statements over correct ones, taismo SEO wiki\" title=\"AI hallucination: when language models invent facts\" width=\"1200\" height=\"675\" style=\"width:100%;height:auto;border-radius:12px\" \/><br \/>\n    <\/figure>\n<h2 id=\"term-profile\">Term profile at a glance<\/h2>\n<table>\n<tr>\n<th>Attribute<\/th>\n<th>Details<\/th>\n<\/tr>\n<tr>\n<td>Part of speech<\/td>\n<td>Noun (the AI hallucination)<\/td>\n<\/tr>\n<tr>\n<td>Plural<\/td>\n<td>AI hallucinations<\/td>\n<\/tr>\n<tr>\n<td>Syllables<\/td>\n<td>AI hal\u00b7lu\u00b7ci\u00b7na\u00b7tion<\/td>\n<\/tr>\n<tr>\n<td>German equivalent<\/td>\n<td>KI-Halluzination<\/td>\n<\/tr>\n<tr>\n<td>Technical synonym<\/td>\n<td>Confabulation<\/td>\n<\/tr>\n<tr>\n<td>Related terms<\/td>\n<td>Large language model, retrieval-augmented generation, grounding, semantic drift<\/td>\n<\/tr>\n<tr>\n<td>Field<\/td>\n<td>Artificial intelligence, AI search, GEO<\/td>\n<\/tr>\n<\/table>\n<p>The term is a metaphor borrowed from psychology. A person who hallucinates perceives something that does not exist. A language model perceives nothing; it produces text that looks like a fact. That is why some experts use the word confabulation, which in medicine describes the unconscious filling of memory gaps.<\/p>\n<h2 id=\"why-models-invent-facts\">Why do language models invent facts?<\/h2>\n<p><strong>Language models invent facts because they are trained on probability and have no built-in stop mechanism for a knowledge gap.<\/strong> A <a href=\"https:\/\/taismo.de\/en\/what-is\/llm\/\">large language model<\/a> predicts word by word what is most likely to come next. An invented year can be just as likely for the model as the correct one, as long as it fits the sentence.<\/p>\n<p>AI hallucinations have 4 main causes:<\/p>\n<ol>\n<li><strong>Probability replaces knowledge:<\/strong> The model stores patterns from its training data. It has no verified fact database, so the most plausible answer wins.<\/li>\n<li><strong>Rare facts stay blurry:<\/strong> A birthday, a company location, or a revenue figure that appears only once in the training material cannot be reconstructed reliably from patterns. This is exactly where the model most often inserts an invented detail.<\/li>\n<li><strong>Evaluation rewards guessing:<\/strong> In September 2025, OpenAI showed in the study &#8220;Why Language Models Hallucinate&#8221; that common benchmarks count only correct answers. A guess has a chance of scoring, while &#8220;I don&#8217;t know&#8221; earns zero points. Models therefore learn to guess.<\/li>\n<li><strong>Errors carry forward:<\/strong> In long answers, every new sentence builds on the previous ones. A detail invented early pulls further invented details along with it that fit logically.<\/li>\n<\/ol>\n<p>The figures from the same OpenAI study show how strongly evaluation matters. On the SimpleQA test, the older model o4-mini answered <strong>24 percent<\/strong> of the questions correctly and <strong>75 percent<\/strong> incorrectly, because it almost never abstained. The newer gpt-5-thinking-mini abstained on 52 percent of the questions, got 22 percent right, and was wrong on only <strong>26 percent<\/strong>. The accuracy is almost the same, while the error rate drops to a third.<\/p>\n<figure>\n      <a href=\"https:\/\/taismo.de\/en\/geo-audit\/\" title=\"GEO audit: trace false AI statements about your brand\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" viewBox=\"0 0 820 340\" role=\"img\" aria-labelledby=\"fig1t fig1d\" style=\"width:100%;height:auto;max-width:820px;font-family:Arial,Helvetica,sans-serif\"><title id=\"fig1t\">How an AI hallucination happens: a backed answer or a plausible invention<\/title><desc id=\"fig1d\">A question goes to the language model, which calculates the most likely continuation. If the model finds a clear source, the result is a backed answer. If it hits a knowledge gap or conflicting information, it picks the most plausible phrasing, and the result is an AI hallucination that sounds confident and is wrong. The orange path shows the route to the hallucination.<\/desc><rect x=\"0\" y=\"0\" width=\"820\" height=\"340\" fill=\"#ffffff\"\/><g opacity=\"0.08\" transform=\"translate(40,222) scale(0.1)\"><path fill=\"#072328\" d=\"M1122.08,481.04l-132.61-85.2h-.19c-2.19-1.43-3.97-3.41-5.16-5.74-2.21-4.37-3.05-9.3-2.42-14.16.76-8.35,4.11-36.09,7.27-61.66,1.56-12.75,3.19-25.06,4.24-34.12s1.88-14.92,1.88-14.95c.16-1.49-.91-2.83-2.4-2.99-.82-.09-1.64.2-2.23.79l-125.88,131.4-8.04,8.39-19.55,17.03c-15.82,12.94-32.67,24.58-50.38,34.79-13.56,10.1-30.33,14.95-47.19,13.65-32.06-1.03-63.83-6.46-94.41-16.13-59.13-27.3-123.67-40.89-188.79-39.76l-154,13.68s-38.52,7.46-47.22-19.9c-17.4-52.51-40.3-103.03-68.33-150.72-18.96-28.44-28.95-61.92-28.7-96.1,0,0,6.22-81.08-38.52-117.11C90.63,17.29,54.43,2.47,16.39,0,6.31,1.22-.93,10.28.1,20.37c.28,4.44,2.08,8.65,5.1,11.92,8.2,9.64,19.73,15.84,32.3,17.38,14.34,3.69,27.2,11.68,36.86,22.89,9.63,12.75,14.99,28.22,15.3,44.19,2.58,35.07,10.68,86.5,23.59,156.23,8.35,25.79,20.49,50.19,36.03,72.41,22.63,40.89,40.55,84.23,53.41,129.16-16.18,27.55-29.5,56.69-39.76,86.95-17.41,51.02-45.98,132.89-45.98,132.89,0,0-45.95,54.65-52.16,62.11-6.22,7.46-23.59,72.03-37.27,90.68-9.6,26.65-13.81,54.93-12.4,83.22,0,0,3.73,28.7,40.97,27.33,37.24-1.37,19.86-36.03,19.86-36.03,0,0-11.16-21.14-6.22-32.3,4.94-11.16,18.65-51.02,18.65-51.02,0,0,63.32-45.95,83.22-57.14,19.9-11.19,125.43-88.19,125.43-88.19,16.5,25.78,22.71,56.81,17.41,86.95-7.46,54.65-1.24,58.38,2.49,70.82,20.08,29.95,42.08,58.58,65.84,85.71,0,0,12.4,21.14,28.7,21.14h42.22s31.88-14.67,12.98-33.57c-14.16-11.24-32.88-14.95-50.25-9.95l-33.54-53.41c-2.51-16.38,1.99-33.06,12.4-45.95,30.61-31.96,50.19-72.88,55.89-116.76l127.92,13.68s26.08,0,23.59,42.22c-2.49,42.22,0,191.31,0,191.31l13.49,24.84h55.89s21.11-32.3-21.14-38.52c1.37-32.24,7.66-64.07,18.65-94.41,7-23.95,17.02-46.91,29.81-68.33,0,0,53.41-39.76,80.73-32.3,27.33,7.46,99.38,82.01,99.38,82.01l100.91,97.92s26.08,36.03,48.46,37.27c13.24.76,26.52.76,39.76,0l9.92-18.65s8.7-26.08-44.64-22.32c0,0-106.81-150.3-119.25-208.65,0,0,.25-18.37-.45-42.82h0l3.99-58.99c9.85-19.32,36.73-45.34,110.16-53.34.49-.25,1-.45,1.53-.61,4.34-1.28,11.45-3.19,17.44-4.91l10.81-2.93c4.03-.96,7.88-2.52,11.45-4.62,9.59-5.47,18.39-12.23,26.15-20.09,10.56-9.11,19.77-19.67,27.36-31.37.61-1.22.19-2.71-.96-3.44M1009.05,488.15c1.73-.31,3.39.85,3.7,2.58.01.08.02.15.03.23h0c.31,1.73-.85,3.39-2.58,3.7-.07.01-.15.02-.23.03l-54.2,7.97-3.76-6.38,57.04-8.13ZM916.33,437.62c2.86-18.87,20.49-31.85,39.36-28.98s31.85,20.49,28.98,39.36c-.55,3.63-1.67,7.14-3.33,10.41-.63,1.34-2.23,1.91-3.56,1.28-.93-.44-1.53-1.37-1.54-2.4-.09-2.7-.78-5.35-2.04-7.75-4.65-8.57-15.36-11.74-23.93-7.09s-11.74,15.36-7.09,23.93c1.3,2.37,3.13,4.4,5.36,5.93,1.23.82,1.56,2.49.73,3.72-.57.85-1.57,1.3-2.58,1.16-2.58-.29-5.11-.86-7.56-1.72l-22.8-37.85ZM1016.77,510.63l-77.06,35.52,29.56-20.85h0l44.64-20.66c1.57-.8,3.49-.18,4.29,1.39.04.08.08.16.11.24h0c.8,1.57.18,3.49-1.39,4.29-.08.04-.16.08-.24.11M1109.14,486.53c-9.52,9.47-20.12,17.77-31.6,24.74-2.9,1.62-5.97,2.91-9.15,3.86-3.58,1.24-7.5,1.04-10.94-.54-2.25-1.23-4.08-3.11-5.26-5.39-2.04-4.33-3.04-9.08-2.93-13.87-.05-2.79.11-5.59.48-8.35,1.44-11.8,2.9-23.58,4.37-35.36.18-1.47,1.52-2.51,2.99-2.32.4.05.78.19,1.12.41l50.47,32.68c.68.44,1.12,1.17,1.21,1.98.09.81-.19,1.62-.76,2.2\"\/><\/g><rect x=\"20\" y=\"130\" width=\"140\" height=\"64\" rx=\"12\" fill=\"#ffffff\" stroke=\"#072328\" stroke-width=\"2\"\/><text x=\"90\" y=\"158\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"16\" font-weight=\"bold\">Question<\/text><text x=\"90\" y=\"178\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"13\">to the model<\/text><path d=\"M162 162 L190 162\" stroke=\"#072328\" stroke-width=\"3\"\/><path d=\"M198 162 L186 156 L186 168 Z\" fill=\"#072328\"\/><rect x=\"204\" y=\"122\" width=\"190\" height=\"80\" rx=\"12\" fill=\"#072328\"\/><text x=\"299\" y=\"152\" text-anchor=\"middle\" fill=\"#ffffff\" font-size=\"15\" font-weight=\"bold\">Language model<\/text><text x=\"299\" y=\"172\" text-anchor=\"middle\" fill=\"#ffffff\" font-size=\"12\">calculates the most<\/text><text x=\"299\" y=\"188\" text-anchor=\"middle\" fill=\"#ffffff\" font-size=\"12\">likely continuation<\/text><path d=\"M396 146 L430 90 L446 90\" fill=\"none\" stroke=\"#072328\" stroke-width=\"3\"\/><path d=\"M454 90 L442 84 L442 96 Z\" fill=\"#072328\"\/><path d=\"M396 178 L430 234 L446 234\" fill=\"none\" stroke=\"#ED8924\" stroke-width=\"3\"\/><path d=\"M454 234 L442 228 L442 240 Z\" fill=\"#ED8924\"\/><rect x=\"460\" y=\"58\" width=\"160\" height=\"64\" rx=\"12\" fill=\"#ffffff\" stroke=\"#072328\" stroke-width=\"2\"\/><text x=\"540\" y=\"86\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"14\" font-weight=\"bold\">Clear source<\/text><text x=\"540\" y=\"106\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"12\">Grounding, RAG, Schema<\/text><rect x=\"460\" y=\"202\" width=\"160\" height=\"64\" rx=\"12\" fill=\"#fdf1e3\" stroke=\"#ED8924\" stroke-width=\"2\"\/><text x=\"540\" y=\"230\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"14\" font-weight=\"bold\">Gap or<\/text><text x=\"540\" y=\"250\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"12\">conflicting information<\/text><path d=\"M622 90 L650 90\" stroke=\"#072328\" stroke-width=\"3\"\/><path d=\"M658 90 L646 84 L646 96 Z\" fill=\"#072328\"\/><path d=\"M622 234 L650 234\" stroke=\"#ED8924\" stroke-width=\"3\"\/><path d=\"M658 234 L646 228 L646 240 Z\" fill=\"#ED8924\"\/><rect x=\"664\" y=\"58\" width=\"140\" height=\"64\" rx=\"12\" fill=\"#ffffff\" stroke=\"#072328\" stroke-width=\"2\"\/><text x=\"734\" y=\"86\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"14\" font-weight=\"bold\">Backed<\/text><text x=\"734\" y=\"106\" text-anchor=\"middle\" fill=\"#072328\" font-size=\"14\" font-weight=\"bold\">answer<\/text><rect x=\"664\" y=\"202\" width=\"140\" height=\"64\" rx=\"12\" fill=\"#ED8924\"\/><text x=\"734\" y=\"230\" text-anchor=\"middle\" fill=\"#ffffff\" font-size=\"14\" font-weight=\"bold\">AI hallucination<\/text><text x=\"734\" y=\"250\" text-anchor=\"middle\" fill=\"#ffffff\" font-size=\"12\">confident and wrong<\/text><text x=\"540\" y=\"290\" text-anchor=\"middle\" fill=\"#55605f\" font-size=\"12\">The model fills the gap with the most plausible phrasing.<\/text><text x=\"20\" y=\"36\" fill=\"#072328\" font-size=\"15\" font-weight=\"bold\">Two paths for the same question<\/text><text x=\"804\" y=\"326\" text-anchor=\"end\" fill=\"#8a9695\" font-size=\"11\">Fig. 1 \u00b7 taismo<\/text><\/svg><\/a><figcaption><strong>Fig. 1: Two paths for the same question. With a clear source, the result is a backed answer; with a gap or conflicting information, the model fills in something plausible.<\/strong><\/figcaption><\/figure>\n<h2 id=\"types\">What types of AI hallucinations are there?<\/h2>\n<p><strong>There are 4 types of AI hallucinations: invented facts, invented sources, contradictions of the source material, and confused entities.<\/strong> Research broadly distinguishes between errors against the world (factuality) and errors against a supplied source (faithfulness).<\/p>\n<table>\n<tr>\n<th>Type<\/th>\n<th>What happens<\/th>\n<th>Typical example<\/th>\n<\/tr>\n<tr>\n<td>Invented facts<\/td>\n<td>The model states figures, dates, or properties that are not true<\/td>\n<td>A wrong founding year or an invented price<\/td>\n<\/tr>\n<tr>\n<td>Invented sources<\/td>\n<td>The model cites studies, rulings, or URLs that do not exist<\/td>\n<td>In Mata v. Avianca, a New York lawyer filed briefs in 2023 containing six court decisions invented by ChatGPT<\/td>\n<\/tr>\n<tr>\n<td>Contradiction of the source<\/td>\n<td>A summary of a document contains statements that are not in the document<\/td>\n<td>A negation gets lost in the condensed version and reverses the statement<\/td>\n<\/tr>\n<tr>\n<td>Confused entities<\/td>\n<td>The model mixes up people or companies with the same or a similar name<\/td>\n<td>The r\u00e9sum\u00e9 of a namesake ends up in the answer about a managing director<\/td>\n<\/tr>\n<\/table>\n<p>For businesses, the first and the fourth type are the most dangerous. An invented source is usually caught during review. A wrong price or a confused <a href=\"https:\/\/taismo.de\/en\/what-is\/entity\/\">entity<\/a>, on the other hand, looks credible and often goes unnoticed until a customer asks about it.<\/p>\n<h2 id=\"semantic-drift\">What is semantic drift and why does the plausible win?<\/h2>\n<p><strong>Semantic drift is the gradual shift in meaning that occurs when a language model summarizes, rephrases, or processes content over several steps.<\/strong> Each individual step picks a phrasing that sounds good and stays close to the original. Taken together, the statement moves away from the source text until it claims something else.<\/p>\n<p>The model prefers the plausible over the correct because that is exactly what it is optimized for. For a language model, plausible means: The phrasing appears frequently in similar texts. Correct means: The statement matches the world. The model can only measure the first from its training data. 3 patterns come up again and again:<\/p>\n<ul>\n<li><strong>Negations disappear:<\/strong> &#8220;Does not issue court-proof expert reports&#8221; becomes &#8220;issues court-proof expert reports&#8221; when condensed, because the affirmative form is more common in everyday language.<\/li>\n<li><strong>The unusual is smoothed into the usual:<\/strong> An unusual figure, such as an 11-month contract term, slips to the industry-standard 12, because that number dominates the training data.<\/li>\n<li><strong>Contradictions get resolved:<\/strong> If the model finds two different values on two pages, it picks the more common one or blends both into a third.<\/li>\n<\/ul>\n<p>The third point is why fact consistency carries so much weight for AI visibility. If a website states 15 years of experience on the homepage, 12 in its company profile, and 10 in a business directory, the model decides on its own which figure applies. If all sources state the same value, there is no gap for the model to fill with something plausible. That is why taismo applies one rule to every website: One answer per fact, sitewide.<\/p>\n<h2 id=\"protect-your-brand\">How do you protect your brand from false AI statements?<\/h2>\n<p><strong>You protect your brand from false AI statements by giving the model a clear, consistent, and easily retrievable answer to every important question.<\/strong> A decision from Canada shows how serious the issue is: In February 2024, Air Canada had to pay a customer a refund that the chatbot on its own website had promised based on an invented rule for bereavement fares. The company was held liable for what its AI said.<\/p>\n<p>5 measures reduce the risk the most:<\/p>\n<ol>\n<li><strong>Standardize your facts:<\/strong> Prices, time frames, headcount, locations, and founding year read the same on every page and in every profile. Every discrepancy is an invitation to guess.<\/li>\n<li><strong>Make the entity unambiguous:<\/strong> <a href=\"https:\/\/taismo.de\/en\/what-is\/structured-data\/\">Structured data<\/a> connects the company, people, and profiles via <code>sameAs<\/code> into a single node in the <a href=\"https:\/\/taismo.de\/en\/what-is\/knowledge-graph\/\">Knowledge Graph<\/a>. If there are namesakes, a <code>disambiguatingDescription<\/code> separates the right person.<\/li>\n<li><strong>Build citable answer pages:<\/strong> A <a href=\"https:\/\/taismo.de\/en\/what-is\/grounding-page\/\">grounding page<\/a> answers the core questions about your business in clear sentences with figures and dates. Systems using <a href=\"https:\/\/taismo.de\/en\/what-is\/retrieval-augmented-generation\/\">retrieval-augmented generation<\/a> pull exactly these passages at answer time.<\/li>\n<li><strong>Align third-party sources:<\/strong> Business directories, review platforms, and press articles feed into the picture a model has of you as <a href=\"https:\/\/taismo.de\/en\/what-is\/brand-mentions\/\">brand mentions<\/a>. Outdated descriptions there can outweigh your own website.<\/li>\n<li><strong>Measure AI answers regularly:<\/strong> Ask ChatGPT, Gemini, Perplexity, and Google AI Mode the same questions about your brand and compare the answers. A structured measurement of AI visibility follows exactly this pattern.<\/li>\n<\/ol>\n<p>A <a href=\"https:\/\/taismo.de\/en\/geo-audit\/\">GEO audit<\/a> narrows down where false statements about your business come from. We check which sources the models cite, where your own information contradicts itself, and which profiles are still spreading an outdated version. One thing to keep in mind: AI systems only pick up corrected information after the next retrieval or training run. A fixed source therefore takes effect with a delay.<\/p>\n<h2 id=\"faq\">Frequently asked questions about AI hallucinations<\/h2>\n<p><strong>Can AI hallucinations be avoided completely?<\/strong><br \/>\nAI hallucinations can be reduced, but by current knowledge they cannot be ruled out entirely. OpenAI also states that the accuracy of a language model will never reach 100 percent, because some questions simply cannot be answered from the available information.<\/p>\n<p><strong>Does ChatGPT hallucinate less with web search?<\/strong><br \/>\nWith web search enabled, ChatGPT usually hallucinates less on factual questions, because it bases its answer on retrieved pages. The risk then shifts to the quality of those pages: If they contradict each other or are outdated, the model adopts the error.<\/p>\n<p><strong>How can you spot an AI hallucination?<\/strong><br \/>\nThe most reliable way to spot an AI hallucination is to check specific details such as figures, names, quotes, and sources against the original source. Red flags are source references without a link, very round numbers, and details that do not appear in any second source.<\/p>\n<\/p><\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>IKI Contents Term profile at a glance Why do language models invent facts? What types of AI hallucinations are there? What is semantic drift and why does the plausible win? How do you protect your brand from false AI statements? Frequently asked questions about AI hallucinations What is an AI hallucination? An AI hallucination is [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":8014,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_taismo_jsonld":"","_taismo_h1":"","taismo_jsonld":"{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/\",\"url\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/\",\"name\":\"What is an AI hallucination?\",\"isPartOf\":{\"@id\":\"https:\/\/taismo.de\/#website\"},\"inLanguage\":\"en-US\",\"primaryImageOfPage\":{\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage\"},\"mainEntity\":{\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#definedterm\"},\"breadcrumb\":{\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#breadcrumb\"},\"datePublished\":\"2026-09-29T07:30:00+02:00\",\"dateModified\":\"2026-09-29T07:30:00+02:00\",\"speakable\":{\"@type\":\"SpeakableSpecification\",\"cssSelector\":[\".brand-claim\",\".lead-def\"]},\"author\":{\"@id\":\"https:\/\/taismo.de\/#person-dominik-breitbach\"},\"publisher\":{\"@id\":\"https:\/\/taismo.de\/#organization\"}},{\"@type\":\"ImageObject\",\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage\",\"contentUrl\":\"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"url\":\"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"width\":1200,\"height\":675,\"caption\":\"AI hallucination: when language models invent facts\"},{\"@type\":\"DefinedTerm\",\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#definedterm\",\"name\":\"AI hallucination\",\"alternateName\":[\"AI hallucinations\",\"confabulation\"],\"termCode\":\"ai-hallucination\",\"url\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/\",\"inLanguage\":\"en-US\",\"inDefinedTermSet\":{\"@id\":\"https:\/\/taismo.de\/#seo-wiki\"},\"description\":\"An AI hallucination is a statement by an AI system that sounds convincing but is factually wrong or not backed by any source. The language model invents facts, figures, quotes, or sources and phrases them with the same confidence as correct information. AI hallucinations happen because a language model calculates the most likely continuation of a text and does not check whether its statement is true.\",\"disambiguatingDescription\":\"In AI, an AI hallucination is a false or unsupported statement by a language model. The term does not refer to a sensory illusion in the medical or psychological sense.\",\"author\":{\"@id\":\"https:\/\/taismo.de\/#person-dominik-breitbach\"},\"publisher\":{\"@id\":\"https:\/\/taismo.de\/#organization\"}},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/taismo.de\/en\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"SEO Glossary\",\"item\":\"https:\/\/taismo.de\/en\/what-is\/\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"AI hallucination\"}]},{\"@type\":\"FAQPage\",\"@id\":\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#faqpage\",\"mainEntity\":[{\"@type\":\"Question\",\"name\":\"Can AI hallucinations be avoided completely?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"AI hallucinations can be reduced, but by current knowledge they cannot be ruled out entirely. OpenAI also states that the accuracy of a language model will never reach 100 percent, because some questions simply cannot be answered from the available information.\"}},{\"@type\":\"Question\",\"name\":\"Does ChatGPT hallucinate less with web search?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"With web search enabled, ChatGPT usually hallucinates less on factual questions, because it bases its answer on retrieved pages. The risk then shifts to the quality of those pages: If they contradict each other or are outdated, the model adopts the error.\"}},{\"@type\":\"Question\",\"name\":\"How can you spot an AI hallucination?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"The most reliable way to spot an AI hallucination is to check specific details such as figures, names, quotes, and sources against the original source. Red flags are source references without a link, very round numbers, and details that do not appear in any second source.\"}}]}]}","footnotes":""},"categories":[80],"tags":[92],"class_list":["post-8029","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-what-is","tag-k-en"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.1 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>What Is an AI Hallucination? Why AI Invents Facts | taismo<\/title>\n<meta name=\"description\" content=\"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"What Is an AI Hallucination? Why AI Invents Facts | taismo\" \/>\n<meta property=\"og:description\" content=\"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0\" \/>\n<meta property=\"og:url\" content=\"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/\" \/>\n<meta property=\"og:site_name\" content=\"taismo| Agentur f\u00fcr SEO &amp; Webdesign\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/TaismoOnlineMarketing\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-29T05:30:00+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"675\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Dominik Breitbach\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/\"},\"author\":{\"name\":\"Dominik Breitbach\",\"@id\":\"https:\\\/\\\/taismo.de\\\/#person-dominik-breitbach\"},\"headline\":\"AI hallucination\",\"datePublished\":\"2026-09-29T05:30:00+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/\"},\"wordCount\":1558,\"publisher\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"keywords\":[\"K\"],\"articleSection\":[\"SEO Glossary\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/\",\"url\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/\",\"name\":\"What Is an AI Hallucination? Why AI Invents Facts | taismo\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"datePublished\":\"2026-09-29T05:30:00+00:00\",\"description\":\"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#primaryimage\",\"url\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"contentUrl\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2026\\\/09\\\/ki-halluzination-warum-ki-fakten-erfindet.webp\",\"width\":1200,\"height\":675,\"caption\":\"KI-Halluzination: Wenn Sprachmodelle Fakten erfinden\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/what-is\\\/ai-hallucination\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Startseite\",\"item\":\"https:\\\/\\\/taismo.de\\\/en\\\/search-marketing\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"AI hallucination\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#website\",\"url\":\"https:\\\/\\\/taismo.de\\\/en\\\/\",\"name\":\"taismo| Agentur f\u00fcr SEO & Webdesign\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#organization\"},\"alternateName\":\"taismo\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/taismo.de\\\/en\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#organization\",\"name\":\"taismo SEO Agentur M\u00fcnchen\",\"alternateName\":\"taismo\",\"url\":\"https:\\\/\\\/taismo.de\\\/en\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/cropped-favicon.png\",\"contentUrl\":\"https:\\\/\\\/taismo.de\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/cropped-favicon.png\",\"width\":512,\"height\":512,\"caption\":\"taismo SEO Agentur M\u00fcnchen\"},\"image\":{\"@id\":\"https:\\\/\\\/taismo.de\\\/en\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/www.facebook.com\\\/TaismoOnlineMarketing\",\"https:\\\/\\\/branchenbuch.portal.muenchen.de\\\/mhp\\\/1380360\\\/\",\"https:\\\/\\\/www.crunchbase.com\\\/organization\\\/taismo-gmbh\",\"https:\\\/\\\/de.trustpilot.com\\\/review\\\/taismo.de\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/taismo\\\/\",\"https:\\\/\\\/www.instagram.com\\\/taismo.de\",\"https:\\\/\\\/onlinemarketing.de\\\/dienstleister\\\/online-marketing-agentur\\\/kirchheim-bei-muenchen\\\/taismo\",\"https:\\\/\\\/www.agenturtipp.de\\\/agentur\\\/taismo-online-marketing\\\/\",\"https:\\\/\\\/startupvalley.news\\\/de\\\/taismo-online-marketing-agentur\\\/\",\"https:\\\/\\\/www.wlw.de\\\/de\\\/firma\\\/taismo-gmbh-1918135\",\"https:\\\/\\\/www.youtube.com\\\/@taismo8632\",\"https:\\\/\\\/techbehemoths.com\\\/company\\\/taismo-gmbh\",\"https:\\\/\\\/www.kununu.com\\\/de\\\/taismo1\",\"https:\\\/\\\/www.provenexpert.com\\\/taismo1\\\/\",\"https:\\\/\\\/www.provenemployer.com\\\/p\\\/taismo-employer\\\/\",\"https:\\\/\\\/www.trustedshops.de\\\/bewertung\\\/taismo-de\",\"https:\\\/\\\/www.kennstdueinen.de\\\/online-marketing-kirchheim-bei-muenchen-taismo-gmbh-d2136613.html\",\"https:\\\/\\\/www.ibusiness.de\\\/dienstleister\\\/jb\\\/5920834fb726.html\",\"https:\\\/\\\/www.seo-united.de\\\/seo-agenturen\\\/taismo-3345\\\/\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"What Is an AI Hallucination? Why AI Invents Facts | taismo","description":"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/","og_locale":"en_US","og_type":"article","og_title":"What Is an AI Hallucination? Why AI Invents Facts | taismo","og_description":"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0","og_url":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/","og_site_name":"taismo| Agentur f\u00fcr SEO &amp; Webdesign","article_publisher":"https:\/\/www.facebook.com\/TaismoOnlineMarketing","article_published_time":"2026-09-29T05:30:00+00:00","og_image":[{"width":1200,"height":675,"url":"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp","type":"image\/webp"}],"author":"Dominik Breitbach","twitter_card":"summary_large_image","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#article","isPartOf":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/"},"author":{"name":"Dominik Breitbach","@id":"https:\/\/taismo.de\/#person-dominik-breitbach"},"headline":"AI hallucination","datePublished":"2026-09-29T05:30:00+00:00","mainEntityOfPage":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/"},"wordCount":1558,"publisher":{"@id":"https:\/\/taismo.de\/en\/#organization"},"image":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage"},"thumbnailUrl":"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp","keywords":["K"],"articleSection":["SEO Glossary"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/","url":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/","name":"What Is an AI Hallucination? Why AI Invents Facts | taismo","isPartOf":{"@id":"https:\/\/taismo.de\/en\/#website"},"primaryImageOfPage":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage"},"image":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage"},"thumbnailUrl":"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp","datePublished":"2026-09-29T05:30:00+00:00","description":"AI hallucinations explained: why language models prefer the plausible over the correct and how to protect your brand from false AI claims. Read on! \ud83e\udde0","breadcrumb":{"@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#primaryimage","url":"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp","contentUrl":"https:\/\/taismo.de\/wp-content\/uploads\/2026\/09\/ki-halluzination-warum-ki-fakten-erfindet.webp","width":1200,"height":675,"caption":"KI-Halluzination: Wenn Sprachmodelle Fakten erfinden"},{"@type":"BreadcrumbList","@id":"https:\/\/taismo.de\/en\/what-is\/ai-hallucination\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Startseite","item":"https:\/\/taismo.de\/en\/search-marketing\/"},{"@type":"ListItem","position":2,"name":"AI hallucination"}]},{"@type":"WebSite","@id":"https:\/\/taismo.de\/en\/#website","url":"https:\/\/taismo.de\/en\/","name":"taismo| Agentur f\u00fcr SEO & Webdesign","description":"","publisher":{"@id":"https:\/\/taismo.de\/en\/#organization"},"alternateName":"taismo","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/taismo.de\/en\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/taismo.de\/en\/#organization","name":"taismo SEO Agentur M\u00fcnchen","alternateName":"taismo","url":"https:\/\/taismo.de\/en\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/taismo.de\/en\/#\/schema\/logo\/image\/","url":"https:\/\/taismo.de\/wp-content\/uploads\/2024\/09\/cropped-favicon.png","contentUrl":"https:\/\/taismo.de\/wp-content\/uploads\/2024\/09\/cropped-favicon.png","width":512,"height":512,"caption":"taismo SEO Agentur M\u00fcnchen"},"image":{"@id":"https:\/\/taismo.de\/en\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/TaismoOnlineMarketing","https:\/\/branchenbuch.portal.muenchen.de\/mhp\/1380360\/","https:\/\/www.crunchbase.com\/organization\/taismo-gmbh","https:\/\/de.trustpilot.com\/review\/taismo.de","https:\/\/www.linkedin.com\/company\/taismo\/","https:\/\/www.instagram.com\/taismo.de","https:\/\/onlinemarketing.de\/dienstleister\/online-marketing-agentur\/kirchheim-bei-muenchen\/taismo","https:\/\/www.agenturtipp.de\/agentur\/taismo-online-marketing\/","https:\/\/startupvalley.news\/de\/taismo-online-marketing-agentur\/","https:\/\/www.wlw.de\/de\/firma\/taismo-gmbh-1918135","https:\/\/www.youtube.com\/@taismo8632","https:\/\/techbehemoths.com\/company\/taismo-gmbh","https:\/\/www.kununu.com\/de\/taismo1","https:\/\/www.provenexpert.com\/taismo1\/","https:\/\/www.provenemployer.com\/p\/taismo-employer\/","https:\/\/www.trustedshops.de\/bewertung\/taismo-de","https:\/\/www.kennstdueinen.de\/online-marketing-kirchheim-bei-muenchen-taismo-gmbh-d2136613.html","https:\/\/www.ibusiness.de\/dienstleister\/jb\/5920834fb726.html","https:\/\/www.seo-united.de\/seo-agenturen\/taismo-3345\/"]}]}},"_links":{"self":[{"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/posts\/8029","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/comments?post=8029"}],"version-history":[{"count":1,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/posts\/8029\/revisions"}],"predecessor-version":[{"id":8030,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/posts\/8029\/revisions\/8030"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/media\/8014"}],"wp:attachment":[{"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/media?parent=8029"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/categories?post=8029"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/taismo.de\/en\/wp-json\/wp\/v2\/tags?post=8029"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}