{"id":30,"date":"2026-09-18T05:30:41","date_gmt":"2026-09-18T05:30:41","guid":{"rendered":"https:\/\/baljit.mankash.com\/blog\/?p=30"},"modified":"2026-09-18T05:52:01","modified_gmt":"2026-09-18T05:52:01","slug":"the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources","status":"publish","type":"post","link":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/","title":{"rendered":"The Oracle\u2019s Paradox"},"content":{"rendered":"\n

How better AI answers can undermine the sources they depend on<\/em><\/p>\n\n\n\n

Imagine an independent reviewer who spends a month testing a new generation of batteries. She buys equipment, repeats measurements, investigates unexpected results, and publishes a careful comparison.<\/p>\n\n\n\n

An AI reads her work alongside several other reviews. When someone asks which battery to buy, it produces a clear answer tailored to their needs. The answer is accurate. The sources are credited. The user gets what they need without opening a single review.<\/p>\n\n\n\n

Nothing in this interaction has to go wrong. The reviewer does good work. The AI does good work. The user saves time.<\/p>\n\n\n\n

But if enough readers stop visiting, and those visits were what funded the testing, who pays for the next round?<\/p>\n\n\n\n

I call this the Oracle\u2019s Paradox<\/strong>: AI can become more useful by reducing our need to visit information sources, while weakening the incentives that keep those sources producing useful information.<\/p>\n\n\n\n

The distinction at its heart is simple:<\/p>\n\n\n\n

\n

AI can make visiting a source unnecessary without making the production of its information unnecessary.<\/p>\n<\/blockquote>\n\n\n\n

This is a paradox of incentives, not a logical contradiction or a prediction of inevitable collapse. It asks whether we can improve the way knowledge is consumed without undermining the conditions under which it is created.<\/p>\n\n\n\n

The work behind an answer<\/h2>\n\n\n\n

There are two different kinds of work in the battery example. The reader needs to find, compare, and understand the results. The reviewer needs to obtain them.<\/p>\n\n\n\n

An excellent summary can remove much of the reader\u2019s work. It does not, by itself, remove the need for someone to run the tests. The cost of delivering an answer and the cost of establishing what is true are different costs.<\/p>\n\n\n\n

This distinction is easy to miss because both activities end in text. A measurement, a discovery, an explanation, and a rewrite can all appear as paragraphs on a screen. Yet their production requirements can be very different. Writing another description of an existing result is not the same task as independently checking whether that result holds.<\/p>\n\n\n\n

In a traffic-funded arrangement, the reader\u2019s visit connects these two activities. Attention helps pay for the work that made the answer possible. In a community forum, the return may instead be recognition, conversation, or the satisfaction of helping someone. A direct relationship with an audience can matter even when no money changes hands.<\/p>\n\n\n\n

Now suppose the answer is delivered elsewhere. The information still reaches the reader, but the revenue, recognition, or participation may no longer reach its producer. The activity that captures the value becomes separated from the activity that bears the cost.<\/p>\n\n\n\n

That separation is not necessarily harmful. A better arrangement could replace it. The problem begins when the old support disappears and nothing sufficient takes its place.<\/p>\n\n\n\n

Nor is every lost webpage a loss of knowledge. A hundred summaries of the same experiment are not a hundred independent experiments. We should care more about losing the experiment\u2014or its only independent replication\u2014than about losing some of the summaries.<\/p>\n\n\n\n

What the evidence tells us<\/h2>\n\n\n\n

There is evidence for parts of this mechanism, although not for the entire chain.<\/p>\n\n\n\n

Pew Research Center analyzed browsing data from 900 U.S. adults in March 2025. Users clicked a traditional search result on 8% of visits to Google pages classified as having an AI summary, compared with 15% of visits to pages without one. Links within the summaries themselves were clicked on in just 1% of visits containing a summary. This was observational research, not a randomized comparison of identical searches, so the difference should not be treated as a clean estimate of AI\u2019s causal effect.[1]<\/a><\/sup><\/p>\n\n\n\n

There is also a counterpoint. In August 2025, Google reported that total organic clicks from Search to websites were relatively stable year over year. That was the company\u2019s own aggregate claim, rather than an independent causal estimate.[2]<\/a><\/sup> The two findings need not contradict each other: fewer clicks per search can coexist with stable total clicks if people search more. Aggregate stability could also conceal substantial changes in which sources receive those visits.<\/p>\n\n\n\n

The production side matters too. A 2024 study in PNAS Nexus<\/em> estimated that, by roughly six months after ChatGPT\u2019s release, posting activity on Stack Overflow had fallen by about 25% relative to comparison platforms. It found no significant change in post quality as measured by peer feedback, and declines were observed among both more- and less-experienced contributors.[3]<\/a><\/sup><\/p>\n\n\n\n

That study suggests a risk beyond advertising revenue: a problem solved privately may no longer generate a public exchange that others can learn from. It does not show that every missing post would have contained valuable new knowledge, or that the same effect occurs everywhere.<\/p>\n\n\n\n

These findings support taking the mechanism seriously. They do not establish that AI is causing an irreversible decline in the world\u2019s information supply. That larger claim would require evidence about what stops being produced, what replaces it, and how the balance changes over time.<\/p>\n\n\n\n

Why the damage could be difficult to see<\/h2>\n\n\n\n

Suppose the incentives for original investigation weaken. The answers do not have to get worse immediately.<\/p>\n\n\n\n

There may still be years of useful work to draw on. Older measurements, explanations, and discussions remain available. An AI could continue giving excellent answers from that accumulated material even as the people who would have updated it reduce their work or leave.<\/p>\n\n\n\n

The gap would become more consequential as the world changes. New products behave differently. Software acquires new bugs. An old finding turns out to depend on conditions that no longer hold. Someone needs to notice, investigate, and make the result available.<\/p>\n\n\n\n

A missing investigation is especially hard to detect. Readers can criticize a bad article. They cannot easily identify an important fact that nobody was funded to discover. The output could look plentiful while the supply of fresh, independent evidence becomes thinner.<\/p>\n\n\n\n

This argument does not require believing that AI is incapable of discovery. An AI could help design an experiment, analyze its results, or operate instruments that collect new observations. If it substantially lowers the cost of producing reliable evidence, that could change the economics for the better.<\/p>\n\n\n\n

But even an AI-run testing operation needs equipment, access, energy, and a reason to perform the test. Predicting how a battery might behave is different from establishing how a particular batch actually behaves. Automating the investigation changes who does the work and what it costs; it does not make the need for evidence disappear.<\/p>\n\n\n\n

Why collapse is not inevitable<\/h2>\n\n\n\n

The strongest objection to the Oracle\u2019s Paradox is straightforward: if original information is valuable to AI companies, why would they allow its supply to disappear?<\/p>\n\n\n\n

They might not.<\/p>\n\n\n\n

Return to the reviewer. Suppose she loses advertising revenue but gains a contract to supply independently verified measurements to several AI services. Her website receives fewer visits, yet her testing is better funded. Users still get convenient answers. The underlying work continues.<\/p>\n\n\n\n

In that case, the funding model has changed rather than failed. There is no reason to force readers back through the old interface merely to preserve it.<\/p>\n\n\n\n

Costs could change as well. A research team might use AI to reduce the effort involved in organizing documents, checking calculations, or preparing a report. If production costs fall sufficiently, a source could remain viable even with lower revenue.<\/p>\n\n\n\n

The risk is therefore conditional: the returns to producing valuable information fall faster than its costs, and replacement support does not close the gap.<\/strong><\/p>\n\n\n\n

The underlying economic tension has precedents. In a 2024 preprint, Jiasun Li applied an analogy with the Grossman\u2013Stiglitz information problem to generative AI: if inexpensive AI output eliminates the incentive to produce fresh information, the AI\u2019s own information supply is undermined. Importantly, the paper uses this tension to argue for coexistence between AI and human creators under its model\u2019s assumptions, rather than complete replacement.[4]<\/a><\/sup><\/p>\n\n\n\n

Coexistence, however, does not settle every concern. A market could support information that is valuable to a few large buyers while leaving less commercial subjects poorly covered. Work could continue behind private contracts but become harder for everyone else to access. A handful of suppliers might thrive while independent alternatives disappear.<\/p>\n\n\n\n

These are different outcomes. Producing less information, producing less independently verified information, and producing information that is no longer publicly accessible are not the same problem. Each deserves to be examined separately.<\/p>\n\n\n\n

Paying for sources is not the same as sustaining knowledge<\/h2>\n\n\n\n

\u201cPay the creators\u201d is a reasonable starting point. It is not a complete design.<\/p>\n\n\n\n

Consider a chain in which a laboratory produces a measurement, a specialist explains it, a publisher distributes the explanation, and an AI synthesizes it for a user. Several participants may contribute genuine value. Paying whichever page the AI retrieves most often does not necessarily fund the activity that made the answer possible.<\/p>\n\n\n\n

We do not need a perfect accounting of every sentence\u2019s contribution before doing anything useful. A contract can fund ongoing testing, a maintained dataset, or a particular area of reporting. The practical question is whether the arrangement supports work that would otherwise be underprovided.<\/p>\n\n\n\n

Payment for automated access is already being explored. In July 2025, Cloudflare introduced a private beta of Pay Per Crawl, designed to let website owners charge AI crawlers for access.[5]<\/a><\/sup> This illustrates a possible way to separate payment for machine access from ordinary readership. It does not establish that the resulting revenue will be sufficient, or that it will reach the original producers of information.<\/p>\n\n\n\n

There is a measurement problem too. A source retrieved once might inform thousands of later answers. A rarely accessed dataset might be indispensable to a small number of important decisions. Retrieval frequency is an imperfect measure of contribution.<\/p>\n\n\n\n

And any payment system creates incentives. Rewarding raw retrieval volume could encourage material designed to attract crawlers rather than to establish something new. Rewarding only established publishers could make it difficult for a new specialist to receive support. These are design risks to address, not reasons to assume compensation cannot work.<\/p>\n\n\n\n

Attribution helps with a different part of the problem. A citation lets readers examine the evidence and recognize its source. But a correct citation does not automatically produce an audience, a subscription, or a payment. An answer can be properly attributed and still leave the underlying work economically unsupported.<\/p>\n\n\n\n

Restricting access can protect a source\u2019s income, but it can also reduce who can learn from or independently check the work. An arrangement that keeps information free for people while charging certain commercial uses might avoid some of that tradeoff. It would still need workable terms, enforcement, and enough demand to fund production.<\/p>\n\n\n\n

No single mechanism has to carry the whole burden. Licensing, direct research contracts, memberships, philanthropy, and lower production costs could support different kinds of work. The test should be what each arrangement sustains, not how closely it resembles the web\u2019s previous business models.<\/p>\n\n\n\n

What we should preserve<\/h2>\n\n\n\n

The wrong ambition would be to preserve every click, every website, or every existing intermediary. People should not have to perform unnecessary reading simply to subsidize a distribution system.<\/p>\n\n\n\n

The valuable things to preserve are the ability to investigate, to test claims independently, to maintain useful records, and to make what is learned available to others.<\/p>\n\n\n\n

That gives us a different way to judge progress. Alongside asking whether an AI produces a better answer, we should ask what supports the production of the next piece of evidence. Does value flow back to that work? Can an independent contributor enter the system? Can others inspect and challenge the result? Are we gaining easier access to knowledge while maintaining the capacity to renew it?<\/p>\n\n\n\n

The Oracle\u2019s Paradox is not that AI makes information sources unnecessary. It is that AI can make them less visible and less rewarded while continuing to depend on what they produce.<\/p>\n\n\n\n

We should be able to enjoy the answer without visiting every source. But the work behind the answer must still be worth doing.<\/strong><\/p>\n\n\n\n

\n\n\n\n

Sources<\/h2>\n\n\n\n

[1]<\/strong> Athena Chapekis and Anna Lieb, \u201cGoogle users are less likely to click on links when an AI summary appears in the results,\u201d<\/a> Pew Research Center, July 22, 2025. The study observed March browsing behavior and reconstructed search results in April; its methodology notes that search results and AI summaries can change over time.<\/p>\n\n\n\n

[2]<\/strong> Liz Reid, \u201cAI in Search is driving more queries and higher quality clicks,\u201d<\/a> Google, August 6, 2025. This is a company statement about its own traffic data.<\/p>\n\n\n\n

[3]<\/strong> R. Maria del Rio-Chanona, Nadzeya Laurentsyeva, and Johannes Wachs, \u201cLarge language models reduce public knowledge sharing on online Q&A platforms,\u201d<\/a> PNAS Nexus<\/em> 3(9), 2024, pgae400. Journal DOI<\/a>. The approximately 25% figure describes the estimated effect near the end of the initial six-month period, not the average decline across the entire period.<\/p>\n\n\n\n

[4]<\/strong> Jiasun Li, \u201cIs Generative AI an Existential Threat to Human Creatives? Insights from Financial Economics,\u201d<\/a> 2024 preprint, arXiv:2407.19586.<\/p>\n\n\n\n

[5]<\/strong> Cloudflare, \u201cIntroducing pay per crawl: Enabling content owners to charge AI crawlers for access,\u201d<\/a> and the July 1, 2025 launch announcement<\/a>.<\/p>\n\n","protected":false},"excerpt":{"rendered":"

How better AI answers can undermine the sources they depend on Imagine an independent reviewer who spends a month testing a new generation of batteries. She buys equipment, repeats measurements, investigates unexpected results, and publishes a careful comparison. An AI reads her work alongside several other reviews. When someone asks which battery to buy, it […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[],"class_list":["post-30","post","type-post","status-publish","format-standard","hentry","category-agentic-ai-systems"],"blocksy_meta":[],"yoast_head":"\nThe Oracle\u2019s Paradox - Baljit Singh<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"The Oracle\u2019s Paradox - Baljit Singh\" \/>\n<meta property=\"og:description\" content=\"How better AI answers can undermine the sources they depend on Imagine an independent reviewer who spends a month testing a new generation of batteries. She buys equipment, repeats measurements, investigates unexpected results, and publishes a careful comparison. An AI reads her work alongside several other reviews. When someone asks which battery to buy, it […]\" \/>\n<meta property=\"og:url\" content=\"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/\" \/>\n<meta property=\"og:site_name\" content=\"Baljit Singh\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-18T05:30:41+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-09-18T05:52:01+00:00\" \/>\n<meta name=\"author\" content=\"baljit@mankash.com\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"baljit@mankash.com\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"10 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/\"},\"author\":{\"name\":\"baljit@mankash.com\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/#\\\/schema\\\/person\\\/4251f556a149c39fac9f2344d3756731\"},\"headline\":\"The Oracle\u2019s Paradox\",\"datePublished\":\"2026-09-18T05:30:41+00:00\",\"dateModified\":\"2026-09-18T05:52:01+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/\"},\"wordCount\":2156,\"articleSection\":[\"Agentic AI Systems\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/\",\"url\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/\",\"name\":\"The Oracle\u2019s Paradox - Baljit Singh\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/#website\"},\"datePublished\":\"2026-09-18T05:30:41+00:00\",\"dateModified\":\"2026-09-18T05:52:01+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/#\\\/schema\\\/person\\\/4251f556a149c39fac9f2344d3756731\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"The Oracle\u2019s Paradox\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/\",\"name\":\"Baljit Singh\",\"description\":\"\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/#\\\/schema\\\/person\\\/4251f556a149c39fac9f2344d3756731\",\"name\":\"baljit@mankash.com\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g\",\"caption\":\"baljit@mankash.com\"},\"sameAs\":[\"https:\\\/\\\/baljit.mankash.com\\\/blog\"],\"url\":\"https:\\\/\\\/baljit.mankash.com\\\/blog\\\/author\\\/baljitmankash-com\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"The Oracle\u2019s Paradox - Baljit Singh","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:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/","og_locale":"en_US","og_type":"article","og_title":"The Oracle\u2019s Paradox - Baljit Singh","og_description":"How better AI answers can undermine the sources they depend on Imagine an independent reviewer who spends a month testing a new generation of batteries. She buys equipment, repeats measurements, investigates unexpected results, and publishes a careful comparison. An AI reads her work alongside several other reviews. When someone asks which battery to buy, it […]","og_url":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/","og_site_name":"Baljit Singh","article_published_time":"2026-09-18T05:30:41+00:00","article_modified_time":"2026-09-18T05:52:01+00:00","author":"baljit@mankash.com","twitter_card":"summary_large_image","twitter_misc":{"Written by":"baljit@mankash.com","Est. reading time":"10 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/#article","isPartOf":{"@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/"},"author":{"name":"baljit@mankash.com","@id":"https:\/\/baljit.mankash.com\/blog\/#\/schema\/person\/4251f556a149c39fac9f2344d3756731"},"headline":"The Oracle\u2019s Paradox","datePublished":"2026-09-18T05:30:41+00:00","dateModified":"2026-09-18T05:52:01+00:00","mainEntityOfPage":{"@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/"},"wordCount":2156,"articleSection":["Agentic AI Systems"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/","url":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/","name":"The Oracle\u2019s Paradox - Baljit Singh","isPartOf":{"@id":"https:\/\/baljit.mankash.com\/blog\/#website"},"datePublished":"2026-09-18T05:30:41+00:00","dateModified":"2026-09-18T05:52:01+00:00","author":{"@id":"https:\/\/baljit.mankash.com\/blog\/#\/schema\/person\/4251f556a149c39fac9f2344d3756731"},"breadcrumb":{"@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/baljit.mankash.com\/blog\/the-oracles-paradox-how-better-ai-answers-can-undermine-their-own-information-sources\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/baljit.mankash.com\/blog\/"},{"@type":"ListItem","position":2,"name":"The Oracle\u2019s Paradox"}]},{"@type":"WebSite","@id":"https:\/\/baljit.mankash.com\/blog\/#website","url":"https:\/\/baljit.mankash.com\/blog\/","name":"Baljit Singh","description":"","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/baljit.mankash.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/baljit.mankash.com\/blog\/#\/schema\/person\/4251f556a149c39fac9f2344d3756731","name":"baljit@mankash.com","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/c01794560ee8f701dc9b6ab721f42ba032aa65e2057906e800fcc73f29fe6bc6?s=96&d=mm&r=g","caption":"baljit@mankash.com"},"sameAs":["https:\/\/baljit.mankash.com\/blog"],"url":"https:\/\/baljit.mankash.com\/blog\/author\/baljitmankash-com\/"}]}},"_links":{"self":[{"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/posts\/30","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/comments?post=30"}],"version-history":[{"count":6,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/posts\/30\/revisions"}],"predecessor-version":[{"id":36,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/posts\/30\/revisions\/36"}],"wp:attachment":[{"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/media?parent=30"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/categories?post=30"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/baljit.mankash.com\/blog\/wp-json\/wp\/v2\/tags?post=30"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}