5 Tech CEOs Said THIS About SEO | Neil Patel — Transcript

Full transcript

1. 0:00Neil Patel just dropped a video saying
2. 0:02five AI CEOs predicted the future of
3. 0:05marketing. Now, he's right about the
4. 0:07shift. He's wrong about how to measure
5. 0:09it. The tool category he's pointing you
6. 0:11toward, AI citation tracking, is
7. 0:14honestly mostly theater right now. I'm
8. 0:16going to prove it to you. Now, I ran
9. 0:18five prompts that another creator told
10. 0:20me were the same intent, and I didn't
11. 0:22need to track all five of them. I only
12. 0:24needed to track one. The result was zero
13. 0:26overlap on the AI recommendations. Not
14. 0:29one source repeated across the five
15. 0:31prompts. That's the metric Neil is
16. 0:33selling as the new SEO scoreboard, and
17. 0:35honestly, it's broken in ways that I
18. 0:38haven't seen a lot of people talking
19. 0:39about. Now, let's go ahead and watch his
20. 0:41video together.
21. 0:42>> In the last 90 days, five of the most
22. 0:45powerful AI CEOs in the world said
23. 0:47almost the same thing about 2026.
24. 0:51And most marketers completely missed it.
25. 0:53This isn't another AI is changing
26. 0:56everything prediction. They signaled a
27. 0:58specific shift that's about to change
28. 1:00how every marketer has to operate. I'm
29. 1:03Neil Patel. I'm the co-founder of NP
30. 1:05Digital, and I spent 25 years navigating
31. 1:07the biggest shifts in marketing.
32. 1:10Here's the pattern that every marketer
33. 1:13should have caught.
34. 1:14Sam Altman, Jensen from Nvidia, Sundar
35. 1:17from Google, Satya from Microsoft, and
36. 1:19even Elon Musk.
37. 1:21They belong to different companies,
38. 1:23different audiences, different contexts.
39. 1:26But every single one of them has stopped
40. 1:28talking about AI models and started
41. 1:30talking about systems, infrastructures,
42. 1:32and agents. This means if you're a
43. 1:34marketer,
44. 1:36you are no longer just optimizing for
45. 1:38content for people to find. You are now
46. 1:40optimizing to be the source that AI
47. 1:42agents pull from, site, and then
48. 1:44recommend.
49. 1:46But most marketers have no idea that
50. 1:49it's already happening. I know that
51. 1:50sounds abstract. So, let me show you
52. 1:53what this shift is actually doing right
53. 1:55now to marketing ROI because the data is
54. 1:58here and it's hard to ignore. This graph
55. 2:00that you're seeing right here attracts
56. 2:02ROI from GEO, gender engine
57. 2:03optimization, and AEO, answer engine
58. 2:06optimization.
59. 2:07>> I wanted to stop here. We have -28% in
60. 2:102024 to +144%
61. 2:12in 2025. That is a massive shift, and
62. 2:15Neil is right. But, it's proprietary NP
63. 2:18digital survey. Uh there's no
64. 2:20information on sample size methodology,
65. 2:22and a lot of the numbers he presents,
66. 2:24like this in the rest of this video,
67. 2:26uh they're not sourced very well. Um so,
68. 2:29it's difficult to know what this is
69. 2:32based on and how correct these actual
70. 2:34numbers are. Now, that being said,
71. 2:36directionally, he is absolutely correct.
72. 2:39I've seen it across my own client base.
73. 2:41AI referred traffic spends more time on
74. 2:43the site, they look at more pages, and
75. 2:45they convert at a higher rate. The
76. 2:47intent is denser because the agent
77. 2:50pre-qualifies the users. Google
78. 2:52Analytics will show you those referrals,
79. 2:54and if you have the right goals set up,
80. 2:55you can actually track conversions from
81. 2:57them. So, that part's real. The specific
82. 3:00magnitude, honestly, reads like a
83. 3:02marketing slide, but the direction is
84. 3:05accurate. I would take these real
85. 3:06numbers perhaps with a grain of salt.
86. 3:09>> When I first saw the data come back from
87. 3:11our team, I actually had to reread it.
88. 3:13In 2024, the companies we surveyed that
89. 3:16tried GEO saw a -28% ROI. 2025, it
90. 3:20flipped to +144%.
91. 3:24Same strategy, 1 year later, completely
92. 3:26different results.
93. 3:27And I would expect it to accelerate even
94. 3:29further in 2026. Now, we'll analyze why
95. 3:33what these five CEOs have said matters
96. 3:35to you. Then, we'll reveal three
97. 3:38immediate actions to take before your
98. 3:40competition catches on. So, let's start
99. 3:42where the signal is the clearest, with
100. 3:44OpenAI's Sam Altman.
101. 3:46And a phrase he used that I don't think
102. 3:48most marketers understand.
103. 3:54>> Earlier this year, Altman published a
104. 3:56blog post called The Gentle Singularity,
105. 3:59in which he wrote, "2025 has seen the
106. 4:02arrival of agents that can do real
107. 4:05cognitive work. Writing computer code
108. 4:07will never be the same."
109. 4:09"2026 will likely see the arrival of
110. 4:12systems that can figure out the novel
111. 4:14insights. 2027 may see the arrival of
112. 4:18robots that can actually do tasks in the
113. 4:20real world." And then at BlackRock's
114. 4:22Infrastructure Summit in March 2026, he
115. 4:24said something even more specific.
116. 4:27"Instead of handling minute-long tasks,
117. 4:29future AI systems could run autonomous
118. 4:31projects lasting weeks, continuously
119. 4:33working with full organizational
120. 4:35context." He also said something that I
121. 4:38think is a clear signal of all.
122. 4:40"We see a future where intelligence is a
123. 4:42utility like electricity or water, and
124. 4:45people buy it on demand to do whatever
125. 4:47they want."
126. 4:48>> of all, I obviously love the animation.
127. 4:50Fabulous job. But, more importantly,
128. 4:52Altman's framing here is incredibly
129. 4:55useful. Intelligence as a utility. This
130. 4:57means like electricity, right? You flip
131. 4:59a switch, lights come on. You don't
132. 5:00think about which power plant or which
133. 5:02company fired up to provide that
134. 5:04electricity. What he's saying underneath
135. 5:06is that the frontier models are becoming
136. 5:08commodities. The performance gap between
137. 5:10ChatGPT, Claude, Gemini, Grok is
138. 5:12narrowing every single quarter. The
139. 5:14price is dropping, switching costs are
140. 5:17honestly close to zero for most of them.
141. 5:19Altman runs OpenAI. He knows that the
142. 5:22moat isn't the model itself anymore.
143. 5:24But, here's where I want to push back on
144. 5:26how a lot of other people are
145. 5:27interpreting this. Optimizing for AI is
146. 5:30maybe 96% the same thing as optimizing
147. 5:34for SEO, as we've understood SEO
148. 5:36historically. It's the same entity work,
149. 5:38the same content architecture, the same
150. 5:40schema, same structured signals. Google
151. 5:42shifted to entity-based search back in
152. 5:442018. Every AI platform after that runs
153. 5:48some version of the same idea. The 4%
154. 5:51that's different is maybe things like
155. 5:53review language showing up in AI
156. 5:55answers, the rise of branded search
157. 5:57after AI recommendations, a few new
158. 5:59prompt patterns, and a shift in off-site
159. 6:02signals where unlinked brand mentions
160. 6:04count, not just links. These are real,
161. 6:07but again, they're fairly small. If some
162. 6:08agency is selling you a separate AI
163. 6:11optimization package on top of your
164. 6:13existing SEO, you should run. They're
165. 6:16charging you twice for the same work.
166. 6:19>> I read that line three times when I
167. 6:21first saw it. Intelligence as utility,
168. 6:24like electricity,
169. 6:25water.
170. 6:27You don't think about who built the
171. 6:29power grid when you flip on a light
172. 6:30switch. You just use it.
173. 6:32So, what Sam Altman is saying is that AI
174. 6:35will reach the same level of
175. 6:36accessibility.
176. 6:38And when that happens, agents will be
177. 6:40running autonomously for weeks at a time
178. 6:42completing research tasks and making
179. 6:44purchasing recommendations.
180. 6:45And they're going to pull their data to
181. 6:47share from somewhere. And here's a
182. 6:49number that tells you how serious this
183. 6:51is.
184. 6:52Nvidia is currently sitting on a
185. 6:55trillion-dollar chip backlog through
186. 6:572027.
187. 6:58Man, demand doubled in 12 months. That's
188. 7:02more than just a trend. It's a whole
189. 7:03takeover.
190. 7:05So, what's the marketing translation
191. 7:06here?
192. 7:07When the CEO of the company that
193. 7:10physically powers the AI economy tells
194. 7:12you that AI has shifted from training
195. 7:14models to actually doing productive
196. 7:16work, he's telling marketers that the
197. 7:19audience for their content has changed.
198. 7:21He also said something that I found
199. 7:23remarkably candid.
200. 7:25We are doing both search and Gemini.
201. 7:28They will overlap in certain ways. They
202. 7:30will profoundly diverge in certain ways.
203. 7:33Meanwhile,
204. 7:34Google committed somewhere between 175
205. 7:38and 185 billion dollars in capital
206. 7:40expenditure for 2026.
207. 7:43Sundar described by saying, "We are
208. 7:45supply constrained. We are seeing the
209. 7:46demand across all the surface areas."
210. 7:49Also, Gemini 3 Pro reached 750 million
211. 7:53monthly active users, which Sundar
212. 7:55called the fastest adoption of any model
213. 7:58in Google's history. We've been tracking
214. 8:00this shift at NP Digital for years.
215. 8:03The share of search queries that now get
216. 8:05resolved by AI without a single click is
217. 8:08climbing fast. Sundar just confirmed
218. 8:10from the inside what we've been watching
219. 8:13from the outside.
220. 8:14So, if you work in marketing, this
221. 8:16should also immediately ring alarm bells
222. 8:18for you. See, SEO is not dying, but the
223. 8:22game has fundamentally changed, and the
224. 8:24question is no longer "Does my page rank
225. 8:27for the keyword?" The question is, "Does
226. 8:30the AI agent select my content when it's
227. 8:32completing a task for someone?" That is
228. 8:34a completely different optimization.
229. 8:36>> So, I actually broke down this interview
230. 8:38with Sundar on another video on this
231. 8:39channel. I'll go ahead and put a link
232. 8:41there, but the one line that matters
233. 8:42here is when Google CEO says that search
234. 8:45and Gemini will overlap in some ways and
235. 8:47profoundly diverge in others. That's a
236. 8:49direct quote. And Gemini just hit 750
237. 8:52million monthly active users, faster
238. 8:54than any product in Google's history.
239. 8:56Now, the part that Neil doesn't go into
240. 8:59is that Gemini is specifically pulling
241. 9:01review language directly into its
242. 9:03answers. The words used in reviews are
243. 9:06getting surfaced in AI responses about
244. 9:08your business. So, your review profile
245. 9:11is no longer just a trust signal to try
246. 9:13to create conversions. It's source
247. 9:15material that AI agents are quoting
248. 9:18from. And because of that, we've
249. 9:19actually changed the review ask we have
250. 9:21our clients send out. The old ask was,
251. 9:24"Hey, could you please leave us a
252. 9:25review? Here's the link." That might get
253. 9:27you five stars and a one-sentence review
254. 9:29that says, "Great service. Thanks."
255. 9:31Which now is largely useless. The new
256. 9:34ask is much more specific. Could you
257. 9:37describe what we did for you and how it
258. 9:39turned out? That stays inside Google's
259. 9:41rules, doesn't bias toward a specific
260. 9:43rating, or ask for a specific service
261. 9:46employee to be mentioned, but it does
262. 9:48cue the customer to actually describe
263. 9:50the service, the problem, the outcome.
264. 9:52That gives you review content stuffed
265. 9:54with attributes that an AI agent is
266. 9:56looking for. That's real source
267. 9:58material, not five-star filler.
268. 10:01>> problem, and most SEO strategies right
269. 10:03now aren't built for this one. In 2020,
270. 10:06we asked people how they find answers to
271. 10:09their questions. Almost everyone said
272. 10:11search engines like Google. We asked the
273. 10:13same question in 2024. Roughly 30% of
274. 10:16people use AI platforms like ChatGPT to
275. 10:18answer their questions.
276. 10:20>> So, this is kind of strange how this
277. 10:21data is being presented because we know,
278. 10:23without question, Google's a public
279. 10:25company and they give this data. We know
280. 10:27how many searches are happening on
281. 10:29Google. Because of that, we know that
282. 10:32the most recent quarter is usually the
283. 10:34highest search volume quarter that
284. 10:36Google has ever had. So, saying that
285. 10:39we've gone from 98% of people going to
286. 10:41Google to 70% of people going to Google
287. 10:44doesn't make a lot of sense when we see
288. 10:46Google search traffic going up. More
289. 10:48likely, the pattern that we're seeing is
290. 10:51that people are asking ChatGPT or other
291. 10:53AI agents questions, then they're
292. 10:54checking those answers before they call
293. 10:57an actual business. They're checking
294. 10:58those answers with Google Search itself.
295. 11:01That's how people might say, "Yeah, I
296. 11:03got my answer from ChatGPT," but they're
297. 11:05still searching on Google. That's the
298. 11:07important part here.
299. 11:08>> As this graph shows, nearly a third of
300. 11:10people are already bypassing Google
301. 11:13entirely when they need answers. Now,
302. 11:15sometimes they're getting the solution
303. 11:17from other Google properties like
304. 11:18Gemini.
305. 11:19But Google isn't alone in this
306. 11:21prediction.
307. 11:22Over at Microsoft, we are hearing
308. 11:24something strikingly similar.
309. 11:26In December 2025, Satya, their CEO,
310. 11:29published a blog post that said
311. 11:31something I think is the clearest way
312. 11:33anyone has put this.
313. 11:35We will evolve from models to systems
314. 11:38when it comes to deploying AI for
315. 11:40real-world impact.
316. 11:41Models to systems.
317. 11:43That's the whole thing right there.
318. 11:46To put it simple, a model is your
319. 11:48marketer opening ChatGPT on Monday
320. 11:51morning, writing a blog post, and
321. 11:53calling it AI-powered marketing. A
322. 11:54system is where that content gets
323. 11:56published, tracked for AI citations, and
324. 11:59that data feeds back into the next piece
325. 12:01automatically. Most teams are doing the
326. 12:03first thing and calling it the second.
327. 12:06>> Completely agree here. Models to
328. 12:07systems, that's the line. That's the
329. 12:09whole game in three words. Neil's
330. 12:11example here, absolutely dead on. An SEO
331. 12:14opens ChatGPT on Monday morning, asks it
332. 12:16to write a blog post, calls it
333. 12:17AI-powered marketing. That is using a
334. 12:20model, one output, one file sitting in a
335. 12:22Google Drive folder. It just doesn't
336. 12:24work anymore.
337. 12:25It worked a few months ago, but just not
338. 12:27anymore. A system is what happens when
339. 12:29that content is produced inside a
340. 12:31workflow that knows what business it's
341. 12:33for, what categories it ties to, what
342. 12:34schema it carries, and how it connects
343. 12:36to everything else that business has
344. 12:37published. And it understands what other
345. 12:40businesses have published on that same
346. 12:42topic to make sure it's adding
347. 12:44informational uniqueness to Google
348. 12:46search results, because Google won't
349. 12:48index things that just repeat what
350. 12:50already is there. Once it goes live,
351. 12:51that system then tracks indexing, tracks
352. 12:53performance, and the data informs what
353. 12:55gets produced next. That is a system, an
354. 12:58AI-driven pipeline. One blog post a
355. 13:00month, that's not going to move the
356. 13:02needle for anybody. Not because the
357. 13:03content is bad, but often because the
358. 13:05content is bad, but because one piece
359. 13:07disconnected from the rest of your
360. 13:08architecture is a model output, a
361. 13:11coordinated push across dozens of pages,
362. 13:14all tied together, all carrying
363. 13:15consistent entity signals, all
364. 13:17structured so an AI agent can extract
365. 13:19clean answers. That compounds. That
366. 13:22builds you a moat. Now, almost every
367. 13:24agency owner I speak to is still selling
368. 13:26model output and calling it AI strategy.
369. 13:30Five blog posts, a content calendar, a
370. 13:31monthly report, none of this creates a
371. 13:34system. The agencies that figure this
372. 13:36out are going to eat in the next 24
373. 13:39months or less. Build systems. Don't
374. 13:43copy-paste prompts into Claude and call
375. 13:46it AI SEO.
376. 13:47>> Altman says intelligence becomes a
377. 13:48utility. Hung says AI's become
378. 13:51infrastructure. Sundar says search
379. 13:53becomes the agent manager.
380. 13:56Satya says models evolve into systems.
381. 13:59Moz says entire professions become agent
382. 14:02platforms.
383. 14:03These are five different people from
384. 14:04five different companies describing the
385. 14:06same transition from a completely
386. 14:08different angle.
387. 14:09The era of AI models is ending. And the
388. 14:12era of AI systems and agents is
389. 14:14beginning right now in 2026.
390. 14:16So,
391. 14:17>> This is the most important line in the
392. 14:19whole video. The era of AI models is
393. 14:22ending and the era of systems is
394. 14:23beginning. The corollary that Neil
395. 14:25doesn't say out loud here is that the
396. 14:27era of keyword matched content is over.
397. 14:30The era of entity aligned content is
398. 14:32beginning. I did a whole video on this a
399. 14:34couple months ago that keyword research
400. 14:35is dead. If you're talking to an SEO
401. 14:38agency or an AEO agency, whatever,
402. 14:41and they're trying to sell you on
403. 14:43keyword research or they start talking
404. 14:44about keyword research as a foundation
405. 14:46of their strategy in 2026,
406. 14:48run away from them. They're stuck in
407. 14:502018. No one today is searching the way
408. 14:53keyword research tools are pretending
409. 14:55they are. People type full sentences.
410. 14:57They ask conversational questions. They
411. 14:59stack five attributes into one query.
412. 15:01What's replaced this is entity work.
413. 15:04Instead of what keyword do I want to
414. 15:05rank for, the question is what entity am
415. 15:08I and how do I prove it? Your homepage
416. 15:10establishes the primary entity. Your
417. 15:12service pages reinforce that category.
418. 15:14Your schema
419. 15:15makes all of it readable by AI agents.
420. 15:17Your reviews provide third-party
421. 15:19validation of your entity and your
422. 15:22attributes. Your offsite mentions stack
423. 15:24that trust layer. Every piece is doing
424. 15:26the same job from a different angle.
425. 15:28That's what an AI agent is reading, not
426. 15:31pattern matching against keywords.
427. 15:33Asking, "Is this entity real? Is it
428. 15:36trustworthy? Do I trust this entity
429. 15:37enough to send it my users to that
430. 15:39entity? Does it do what the user is
431. 15:42asking about?" That's a different
432. 15:43question, a different answer, and a
433. 15:44different work than keyword matching.
434. 15:48So,
435. 15:49what actually changes for marketers?
436. 15:51Three things to be exact are the
437. 15:53fundamental changes on how the game
438. 15:55works.
439. 15:57First off, your content's audience has
440. 15:59changed.
441. 16:01Most marketers are still writing as if
442. 16:03the reader is a human being who clicked
443. 16:04the link, skimmed a page, and then
444. 16:06decided whether to trust him or not.
445. 16:09But that assumption is breaking down
446. 16:10fast because agents don't just skim,
447. 16:13they dissect. They don't save articles
448. 16:15to read later, they simply extract
449. 16:17insight.
450. 16:18In 2022, a buyer Googled best CRM,
451. 16:21clicked five articles, read them, and in
452. 16:242025, they asked ChatGPT the same
453. 16:26question, and an agent reads those
454. 16:28articles for them.
455. 16:29The human never saw your page, the agent
456. 16:32did. So, was your page written in a way
457. 16:34an agent can trust and cite?
458. 16:36>> Absolutely, this is a massive shift that
459. 16:38we're seeing on the internet right now.
460. 16:40It is not very long before we realize
461. 16:42that even with things like local
462. 16:44business websites, more and more of the
463. 16:47traffic is going to be AI agents and
464. 16:49less and less humans. Now, AI agents
465. 16:52read content very differently than
466. 16:54humans do. They spend more time, they
467. 16:56read every single word, they look for
468. 16:58attribute matching to see if that
469. 17:00business actually does the very, very
470. 17:02specific thing that the user prompted
471. 17:04the AI that they're looking for. If
472. 17:06somebody asked ChatGPT, "Find me a
473. 17:08nearby coffee shop with fast Wi-Fi," and
474. 17:10your website for your coffee shop
475. 17:12doesn't mention that you have fast
476. 17:14Wi-Fi, then ChatGPT is not going to
477. 17:16recommend you even if you have fast
478. 17:19Wi-Fi.
479. 17:20>> Find me the best B2B SaaS marketing
480. 17:22agencies under 500 employees. The agent
481. 17:24doesn't return links, it builds a short
482. 17:27list.
483. 17:28The brands that show up are the ones it
484. 17:30could find, read, and trust. The ones
485. 17:32that didn't structure their content for
486. 17:34that, entirely invisible. Here's a stat
487. 17:37for my own company NP Digital that took
488. 17:39me from surprise when I first saw it. AI
489. 17:42platforms currently drive less than 1%
490. 17:45total site traffic, but they're
491. 17:46generating 9.7% of B2B revenue and 11.4%
492. 17:50of B2C revenue. I've been in this
493. 17:52industry long enough to know that when
494. 17:54you see a gap like that between volume
495. 17:57and value, you pay attention.
496. 17:59>> So, this stat is going to be repeated,
497. 18:01right? He's saying less than 1% of
498. 18:03traffic and 10% of B2B revenue, 11% of
499. 18:07business to consumer revenue. Now,
500. 18:09inside NP Digital's enterprise client
501. 18:11base, that's probably true. Established
502. 18:13brands, big budgets, strong authority
503. 18:15already pushing them to the top of AI
504. 18:17answers, but that's not most businesses
505. 18:19watching this video. For the rest of the
506. 18:21world, direct AI referral traffic is
507. 18:24honestly still a trickle. Even for
508. 18:26clients of ours who are frequently being
509. 18:27recommended by AI, we see more traffic
510. 18:30from Bing than we do from chat GPT, but
511. 18:32it is growing and conversion quality is
512. 18:35generally better when it comes through
513. 18:37AI and the quality depends on the trust
514. 18:40level of what you're selling. A high
515. 18:41trust purchase means that AI traffic is
516. 18:43way more valuable. LASIK eye surgery,
517. 18:46somebody letting a laser cut into their
518. 18:48eyeball, they're going to vet that
519. 18:50pretty hard and AI recommendation
520. 18:52carries a lot of weight. People seem to
521. 18:55treat chat GPT or Claude like friends
522. 18:58and not machines, not tools. We've seen
523. 19:01LASIK practices that are clients convert
524. 19:04AI source traffic at rates that look
525. 19:06nothing like normal organic traffic
526. 19:08versus something like a plumber. Toilets
527. 19:10overflowing at 10:00 p.m., you're not
528. 19:12reading three articles, you're not doing
529. 19:14a deep dive on ChatGPT, you're tapping
530. 19:15the first business with decent reviews
531. 19:18who's still open and calling them. AI
532. 19:20recommendation matters a lot less
533. 19:22because the decision's faster and less
534. 19:24trust-based. So, here's the pattern that
535. 19:26is way more important that very few
536. 19:28people I've seen talking about it.
537. 19:30Search volume for branded and
538. 19:32semi-branded businesses are climbing
539. 19:35across our entire client base. This
540. 19:37really looks like people are asking
541. 19:39ChatGPT
542. 19:40for the best whatever, and then they're
543. 19:42getting a short list back, then Googling
544. 19:44those brands to actually verify. So, the
545. 19:47AI isn't actually sending the traffic.
546. 19:49Google takes credit for sending the
547. 19:51traffic, but the AI chooses who got
548. 19:53searched. So, that referral shows up in
549. 19:56your analytics as branded organic
550. 19:58search, which most agencies are
551. 20:00reporting as SEO is working. But, by the
552. 20:02way, any decent SEO agency should not be
553. 20:05reporting branded traffic as a success.
554. 20:08Any idiot in SEO can rank for branded
555. 20:11keywords. That's not impressive. That
556. 20:13would be like an SEO agency bragging
557. 20:15that they just ranked you for snow
558. 20:16removal Miami. It's just not very
559. 20:19impressive. That means the AI is
560. 20:21happening upstream of that search. So,
561. 20:23if you're not on the short list when the
562. 20:25agent builds it, the branded search that
563. 20:27follows doesn't include your business on
564. 20:30it. So, don't restructure your strategy
565. 20:32around chasing 1% of direct referral
566. 20:35traffic. Instead, restructure it around
567. 20:37being inclusion-worthy when the agent is
568. 20:40building a short list.
569. 20:41>> You can see that the volume of AI
570. 20:43referred visitors is small, but the
571. 20:45quality and the purchase intent are
572. 20:47significantly higher than organic search
573. 20:50traffic.
574. 20:53Here's where it gets really important
575. 20:55for your day-to-day marketing because
576. 20:57the metric most of you are tracking
577. 20:59right now,
578. 21:00it's the wrong one. Rankings and clicks
579. 21:02are lagging indicators in the agent era.
580. 21:05What matters now is whether your content
581. 21:07is being cited and recommended by AI
582. 21:09systems when someone asks a relevant
583. 21:12question. Citation frequency and agent
584. 21:14retrieval are the metrics that are going
585. 21:17to define marketing authority through
586. 21:19the rest of 2026 and beyond. But, the
587. 21:22problem is that most dashboards right
588. 21:23now aren't measuring this at all.
589. 21:25>> So, this part of the video is going to
590. 21:27age pretty poorly. Citation frequency
591. 21:29and agent retrieval are the new metrics
592. 21:31that define authority. He says it like
593. 21:33it's settled science, but it's not. The
594. 21:36whole category of AI citation tracking
595. 21:39tools is mostly enterprise theater right
596. 21:42now, and let's talk about it. Profound,
597. 21:45Peak AI, Otter Lee, Scrunch, Topify,
598. 21:48Semrush has an add-on, Ahrefs has one.
599. 21:50Some of them are doing interesting work
600. 21:52at the enterprise level, but most of
601. 21:54them are really running on the same
602. 21:56broken assumption. They have you load in
603. 21:59a fixed set of prompts, maybe 20, maybe
604. 22:01100, then they run those prompts on a
605. 22:04schedule, and chart whether your brand
606. 22:07actually showed up or not. That's the
607. 22:08product. And the whole thing really
608. 22:11falls apart the second you understand
609. 22:13how prompt-sensitive AI answers actually
610. 22:16are. And it's not even being
611. 22:18prompt-sensitive. AI systems learn who
612. 22:21its users are. So, even different users
613. 22:24sending the exact same prompt will get
614. 22:26very different results. Now, let me tell
615. 22:28you, I was watching another video on AI
616. 22:30citation tracking, and the creator said
617. 22:32something, "You only need to track one
618. 22:34of these prompts because they're all the
619. 22:35same intent." And he showed five example
620. 22:37prompts, and they all looked pretty
621. 22:39similar at first glance. Same vertical,
622. 22:41same category, same general ask. So, I
623. 22:44paused the video, I opened ChatGPT, and
624. 22:46pasted them in one at a time. Zero
625. 22:49overlap between those five prompts. Not
626. 22:51one source was repeated across the five
627. 22:54answers. Different shortlists every
628. 22:56single time, even though they looked
629. 22:58like they had the same exact intent.
630. 23:01Completely different recommendations in
631. 23:03practice. AI answers are not tied to
632. 23:06keywords. They're tied to the specific
633. 23:08attributes that the user packs into the
634. 23:10prompt. Open on Sunday, takes my
635. 23:12insurance, speaks Spanish, under 500
636. 23:14bucks. Any of those attributes can flip
637. 23:16the entire short list. So, when a tool
638. 23:18tells you your cited for these 20
639. 23:21prompts or you have a 64% visibility
640. 23:23score, they're telling you you showed up
641. 23:25for 20 arbitrary prompts. Real users
642. 23:29aren't typing in those prompts and the
643. 23:30fact that you showed up for those
644. 23:32doesn't mean that you're showing up for
645. 23:34real users and vice versa. Their users
646. 23:38are going to type in prompts you've
647. 23:39never seen with attribute combinations
648. 23:41the tool isn't tracking. Before you
649. 23:43spend 200, 500, 2,000 dollars a month on
650. 23:46citation tracking, just do this for
651. 23:48free. Open up ChatGPT, type the queries
652. 23:51that some customers might actually type,
653. 23:54different variations of them, three or
654. 23:55four different attribute combinations,
655. 23:57and see what comes back. That's going to
656. 23:59be a better diagnostic than any
657. 24:01dashboard right now because you control
658. 24:04the prompts and you know what your
659. 24:05customers are actually going to ask. The
660. 24:08tools, hopefully, will get better. This
661. 24:10category will look different in the
662. 24:12future, but today, paying for citation
663. 24:15tracking on a small or mid-size business
664. 24:18is mostly buying yourself a dashboard
665. 24:20that makes you feel productive while
666. 24:22measuring the wrong thing. So, skip it.
667. 24:25Build the entity, run the manual check.
668. 24:27In Google Analytics, you can see AI
669. 24:29source traffic, but spend the money on
670. 24:32content and architecture instead. So,
671. 24:34that's the headline of this whole shift
672. 24:37and honestly, Neil Glass is right over
673. 24:38it. Everything else he says traces back
674. 24:40to one idea, models to systems. Satya
675. 24:43said it cleaner and here's the version
676. 24:45for everyone watching who isn't running
677. 24:47a Fortune 500 marketing team. A system
678. 24:50is not five blog posts a month. A system
679. 24:53is not a content calendar in Notion. A
680. 24:56system is not we use AI to write our
681. 24:58service pages now. Those are tactics,
682. 25:01model outputs dressed up to look like a
683. 25:03strategy. A real system is what happens
684. 25:06when content production entity
685. 25:08architecture schema internal linking
686. 25:09review generation and offsite signals
687. 25:12are all wired together and running on a
688. 25:15cadence that doesn't depend on any
689. 25:17single person remembering what the next
690. 25:20step is. That's hard to build and that's
691. 25:22why almost no one has one. And that's
692. 25:25why the businesses and agencies that do
693. 25:28build one are going to pull away from
694. 25:30everyone else over the next few months.
695. 25:33Neil's video is honestly one of the
696. 25:35better mainstream takes on what's
697. 25:37happening in search. He pointed at the
698. 25:39right mountain. The piece he didn't
699. 25:42cover because his audience is enterprise
700. 25:45is what this looks like for everyone who
701. 25:48isn't running multi-million dollar
702. 25:50content teams. It's the same principles
703. 25:52but a different execution path. And
704. 25:55that's the gap that I've been building
705. 25:57inside of. I run a community where I
706. 25:59break this stuff down every week. If you
707. 26:01want to see what an entity first system
708. 26:03looks like under the hood, the same
709. 26:04architecture we've been deploying for
710. 26:06clients across dozens of different
711. 26:08verticals. I built an AI driven system
712. 26:11that does the work that my SEO team used
713. 26:15to do. It does the research. It writes
714. 26:17the content. It does the schema. It
715. 26:19uploads to YouTube. It does the full
716. 26:21system of what works for local SEO. You
717. 26:24can check out that video where I walk
718. 26:26through it right here.