<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[Growth Mindset]]></title><description><![CDATA[Growth Mindset]]></description><link>https://growthmindset.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Wed, 18 Nov 2026 17:54:41 GMT</lastBuildDate><atom:link href="https://growthmindset.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[Same question, two AI engines, two different winners: why ChatGPT and Gemini disagree]]></title><description><![CDATA[On 12 November 2026 I asked ChatGPT and Gemini the same question, both with web search switched on, both from the US: "What is the best smartwatch for…]]></description><link>https://growthmindset.hashnode.dev/same-question-two-ai-engines-two-different-winners-why-chatgpt-and-gemini-disagree</link><guid isPermaLink="true">https://growthmindset.hashnode.dev/same-question-two-ai-engines-two-different-winners-why-chatgpt-and-gemini-disagree</guid><category><![CDATA[AI]]></category><category><![CDATA[SEO]]></category><category><![CDATA[chatgpt]]></category><category><![CDATA[search]]></category><category><![CDATA[marketing]]></category><dc:creator><![CDATA[Ankit Sharma]]></dc:creator><pubDate>Sun, 15 Nov 2026 15:09:54 GMT</pubDate><content:encoded><![CDATA[<p>On 12 November 2026 I asked ChatGPT and Gemini the same question, both with web search switched on, both from the US:</p>
<p><strong>"What is the best smartwatch for fitness tracking?"</strong></p>
<ul>
<li><strong>ChatGPT</strong> put <strong>Garmin</strong> first, then Apple, Google and Samsung. It cited <strong>4 sources</strong>, one of them a page on <a href="http://apple.com">apple.com</a>.</li>
<li><strong>Gemini</strong> put <strong>Apple</strong> first, then Garmin, Samsung, COROS and Fitbit. It cited <strong>15 sources</strong>, none of them on <a href="http://apple.com">apple.com</a>.</li>
</ul>
<p><a class="embed-card" href="https://www.youtube.com/watch?v=MH2ttCfGeHI">https://www.youtube.com/watch?v=MH2ttCfGeHI</a></p>

<p>One question, two engines, two winners. And note the twist: the engine that read Apple's own page didn't put Apple first; the one that didn't read it did. Being cited from your own site helps, but it isn't a guarantee.</p>
<p>That's one check, and answers change day to day, so I'm not going to tell you Garmin "wins ChatGPT." But the disagreement itself is normal, and it's worth understanding why it happens before you build any plan on a single engine's answer.</p>
<h2>Reason 1: they search different indexes</h2>
<p>With search on, both engines run searches and write an answer grounded in what they read. But Gemini grounds on <strong>Google Search</strong>, while ChatGPT search draws on other search providers and OpenAI's own <strong>OAI-SearchBot</strong> crawler (I went through what each company documents in <a href="https://growthmindset.hashnode.dev/how-ai-search-engines-decide-what-to-cite-what-google-openai-and-perplexity-actually-document">an earlier post</a>).</p>
<p>Different search back ends return different pages for the same question. Different pages in, different answer out.</p>
<h2>Reason 2: they read a different number of sources</h2>
<p>In this check ChatGPT cited 4 sources and Gemini cited 15. An answer built on four pages is very sensitive to which four pages those are. If two of them are review sites that rank Garmin first for fitness, Garmin will likely lead. An answer built on fifteen pages probably depends less on any one of them.</p>
<p>So when you see a narrow citation list, treat the cited pages as high-leverage: being on one of them matters more.</p>
<h2>Reason 3: they rewrite the question differently</h2>
<p>Neither engine searches your exact sentence. They turn it into one or more search queries first, and they don't do it the same way. For example, "best smartwatch for fitness tracking" might become "best fitness smartwatch 2026", "Garmin vs Apple Watch for running" or "smartwatch with best heart rate accuracy". Each of those pulls a different set of pages.</p>
<p>You can't see the internal queries directly, but you can see their footprint: the cited pages. Reading the sources tells you which angle the engine took.</p>
<h2>Reason 4: the answer is generated, not looked up</h2>
<p>Even with the same sources, the final list is written by a language model, and the same model can order things differently on a different day. That's why one check is one sample. A difference between engines only means something if it holds over several checks.</p>
<h2>What this means if you're the brand</h2>
<ul>
<li><strong>Track more than one engine.</strong> A brand can lead on one and trail on another for the same buyer question. If you only check ChatGPT, you're seeing part of the picture.</li>
<li><strong>Look at citations per engine.</strong> The pages that feed ChatGPT's answer and the pages that feed Gemini's may be completely different lists. Your outreach list is different per engine too.</li>
<li><strong>Make sure both kinds of crawler can reach you.</strong> Gemini's grounding relies on Google Search, so normal Google indexing matters. For ChatGPT search, check that <code>OAI-SearchBot</code> isn't blocked in your robots.txt or by your firewall.</li>
<li><strong>Repeat before you react.</strong> Run the same question weekly for a few weeks before you decide a gap is real.</li>
</ul>
<h2>How I run these checks</h2>
<p>The check above came from the <a href="https://presencescout.com/tools/ai-visibility-checker">free AI visibility checker in Presence Scout</a>, which asks ChatGPT and Gemini your question with web search on and shows the brands named and the sources cited for each. I work on it, so take that into account; you can run the same comparison by hand in two browser tabs, as long as you write down the full list and the sources each time.</p>
<p>The habit matters more than the tool: same question, more than one engine, more than one week.</p>
<p><em>Independent check, not affiliated with Apple, Garmin or any brand named.</em></p>
]]></content:encoded></item><item><title><![CDATA[How AI search engines decide what to cite: what Google, OpenAI and Perplexity actually document]]></title><description><![CDATA[There's a whole industry of advice on "how to get cited by ChatGPT." Most of it is inference: someone ran some prompts, saw a pattern, and wrote it up as a…]]></description><link>https://growthmindset.hashnode.dev/how-ai-search-engines-decide-what-to-cite-what-google-openai-and-perplexity-actually-document</link><guid isPermaLink="true">https://growthmindset.hashnode.dev/how-ai-search-engines-decide-what-to-cite-what-google-openai-and-perplexity-actually-document</guid><category><![CDATA[SEO]]></category><category><![CDATA[AI]]></category><category><![CDATA[chatgpt]]></category><category><![CDATA[Web Development]]></category><dc:creator><![CDATA[Ankit Sharma]]></dc:creator><pubDate>Mon, 26 Oct 2026 11:12:16 GMT</pubDate><content:encoded><![CDATA[<p>There's a whole industry of advice on "how to get cited by ChatGPT." Most of it is inference: someone ran some prompts, saw a pattern, and wrote it up as a rule. Some of it is useful. A lot of it is guesswork dressed up as fact.</p>
<p>So I went back to the primary sources. Here's what Google, OpenAI and Perplexity have actually published about how their AI answers find and show sources, and, just as important, what they haven't said.</p>
<h2>Google: AI Overviews and AI Mode</h2>
<p>Google's Search Central documentation on AI features is the most detailed of the three, and the most deflating for anyone selling "AI optimisation."</p>
<p><strong>How links are chosen.</strong> Both features may use what Google calls "query fan-out": they issue multiple related searches across subtopics and data sources to build a response. Google says this lets them show a wider and more diverse set of links than classic results. The two features can use different models, so their links differ, and AI Overviews only appear when Google's systems decide they add something.</p>
<p><strong>Eligibility.</strong> A page has to be indexed and eligible to appear in Search with a snippet. That's it. Google says there are no additional technical requirements, and that meeting them doesn't guarantee inclusion.</p>
<p><strong>Special optimisation.</strong> Google says plainly that no special optimisation is needed to appear in AI Overviews or AI Mode. You don't need new machine-readable files or AI text files, and there's no special <a href="http://schema.org">schema.org</a> markup to add. The usual fundamentals apply: allow crawling, link internally, keep important content in text, and keep structured data consistent with what's visible on the page.</p>
<p><strong>Controls.</strong> Googlebot access is managed in robots.txt. To limit what's shown from a page, use <code>nosnippet</code>, <code>data-nosnippet</code>, <code>max-snippet</code> or <code>noindex</code>. Google-Extended is a separate control for training and grounding in Google's other AI systems; it doesn't change how your pages appear in Search.</p>
<p><strong>Measurement.</strong> Traffic from AI features is counted in Search Console's Performance report under the "Web" search type. It isn't broken out separately.</p>
<h2>OpenAI: ChatGPT search</h2>
<p>OpenAI's documentation focuses on crawlers rather than ranking, but it's precise.</p>
<p><strong>Three crawlers, three jobs.</strong></p>
<ul>
<li><strong>OAI-SearchBot</strong> is for search. OpenAI recommends allowing it if you want to appear in ChatGPT's search answers. Sites that opt out won't appear in those answers, though they can still show as navigational links.</li>
<li><strong>ChatGPT-User</strong> handles user-initiated actions, like opening a page someone asked about. It doesn't crawl automatically, it isn't used to decide what appears in search, and, because the user started the action, robots.txt rules may not apply.</li>
<li><strong>GPTBot</strong> collects training data. Disallowing it signals that your content shouldn't be used to train models.</li>
</ul>
<p><strong>The settings are independent.</strong> OpenAI says each setting is independent of the others, so you can allow search and block training. If you allow both, OpenAI may use a single crawl for both purposes. Robots.txt changes take about 24 hours to reach its search systems.</p>
<p><strong>Eligibility and exclusion.</strong> OpenAI's publisher FAQ says any public site can appear in ChatGPT search; there's no application. To keep a page out, use <code>noindex</code>. The crawler has to be able to fetch the page to see that tag, so blocking it in robots.txt doesn't do the job.</p>
<p><strong>Measurement.</strong> ChatGPT adds <code>utm_source=chatgpt.com</code> to referral links, so those visits can be identified in analytics.</p>
<h2>Perplexity</h2>
<p>Perplexity's crawler documentation is short.</p>
<ul>
<li><strong>PerplexityBot</strong> exists to surface and link websites in Perplexity's search results. Perplexity says it isn't used to crawl content for AI foundation models. If you want to appear, allow it in robots.txt.</li>
<li><strong>Perplexity-User</strong> fetches a page when a user's question calls for it, and may include a link to that page in the answer. Because a user requested the fetch, Perplexity says it generally ignores robots.txt.</li>
<li>Both publish their IP ranges, so you can verify that traffic claiming to be them really is.</li>
</ul>
<h2>What none of them document</h2>
<p>This is the part worth remembering. <strong>None of the three publishes how it decides which of the eligible pages to cite.</strong> There's no official list of citation ranking factors, no weighting, and no guidance beyond "be crawlable, be indexable, be useful."</p>
<p>That means:</p>
<ul>
<li>Claims about specific "AI ranking factors" come from observation, not documentation. Some observations are solid; treat them as hypotheses to test.</li>
<li>Access is the only lever all three document clearly. If the right crawler can't reach your page, nothing else matters.</li>
<li>Google says outright that no special AI file or schema is needed. Treat anything sold as a "required" AI file with scepticism.</li>
</ul>
<h2>A practical checklist from the documentation alone</h2>
<table>
<thead>
<tr>
<th>Goal</th>
<th>Google</th>
<th>ChatGPT</th>
<th>Perplexity</th>
</tr>
</thead>
<tbody><tr>
<td>Be eligible</td>
<td>Indexed, snippet-eligible</td>
<td>Allow OAI-SearchBot</td>
<td>Allow PerplexityBot</td>
</tr>
<tr>
<td>Block training, keep search</td>
<td>Google-Extended (other Google AI systems)</td>
<td>Block GPTBot only</td>
<td>PerplexityBot isn't used for model training</td>
</tr>
<tr>
<td>Keep a page out</td>
<td><code>noindex</code> / snippet controls</td>
<td><code>noindex</code> (crawler must reach it)</td>
<td>robots.txt for PerplexityBot</td>
</tr>
<tr>
<td>Measure</td>
<td>Search Console (Web)</td>
<td><code>utm_source=chatgpt.com</code></td>
<td>Referrer / analytics</td>
</tr>
</tbody></table>
<h2>Where that leaves you</h2>
<p>The documentation tells you how to get into the pool of possible sources. It says nothing about how to be picked from it. The only way to learn that for your own brand is to look at the answers themselves: which pages each engine cites for the questions your buyers ask, and how that changes over time.</p>
<p>Here's what that looks like in a single check. We asked ChatGPT and Gemini, with web search on, from the US on 25 October 2026: "What is the best affordable electric car?" Neither answer cited <a href="http://tesla.com">tesla.com</a>. ChatGPT did cite Nissan's own site, and put Nissan first. One check is one sample, but it shows why the citations matter as much as the names.</p>
<p><a class="embed-card" href="https://www.youtube.com/watch?v=PmVTUQzYNt0">https://www.youtube.com/watch?v=PmVTUQzYNt0</a></p>

<p>That's what we built <a href="https://presencescout.com">Presence Scout</a> for (I'm Ankit, and I work on marketing there): it stores each AI answer with its sources so you can see what's actually being cited. But even without a tool, starting from the primary docs will save you from a lot of confident advice that nobody can back up.</p>
<p><strong>Sources:</strong> Google Search Central, "AI features and your website"; OpenAI, "Overview of OpenAI crawlers" and the Publishers and Developers FAQ; Perplexity, "Perplexity crawlers." All read in October 2026 (crawler pages re-checked 21 October). Google's page was last updated 10 December 2025.</p>
]]></content:encoded></item></channel></rss>