"Your Website Won't Matter in 3 Years" — Transcript

Full transcript

1. 0:00your website won't matter in three years
2. 0:02or that's the claim. AI agents will find
3. 0:04the business, compare options, and
4. 0:06eventually book the appointment without
5. 0:07the customer ever needing to visit a
6. 0:09website. I think that future is probably
7. 0:12coming, but I'm not sure I agree the
8. 0:14conclusion of websites not matter
9. 0:16follows from it. Let's watch.
10. 0:18>> The CEO of Google, Sundarbachai, said in
11. 0:20a recent interview that 2027 is the year
12. 0:22of the agents. At the same time, 58% of
13. 0:25Google searches end with no clicks. That
14. 0:27means over half the people that are
15. 0:28going to Google and searching for things
16. 0:30never end up on a website. They're
17. 0:31getting their answers from AI. In this
18. 0:33video, I'll show you why websites will
19. 0:34matter dramatically less over the next 3
20. 0:36years.
21. 0:36>> So that 58% number is actually a little
22. 0:39out of date. The newer data from earlier
23. 0:412026 puts zeroclick searches at about
24. 0:4468%. So if anything, the trend he's
25. 0:46pointing to is stronger now than when
26. 0:48that original study came out. And Google
27. 0:51is in fact very clearly moving in this
28. 0:52direction. Sundar Pachai has been
29. 0:55talking about search becoming more and
30. 0:56more agentic. I watched his interview on
31. 0:59this channel earlier. Google is already
32. 1:01rolling out agents inside search that
33. 1:03can reason across information on the
34. 1:05internet and eventually take actions for
35. 1:07you. So the basic premise here, I'm on
36. 1:09board. All right, before we talk about
37. 1:10why websites are not going to be
38. 1:12important, let's talk about where we are
39. 1:13right now. So essentially before people
40. 1:15were searching for keywords. They were
41. 1:17searching for Hback repair near me and
42. 1:19they were finding businesses. They were
43. 1:21reading reviews and then they were going
44. 1:22to websites and converting into leads.
45. 1:24Exactly what I just showed you. Now
46. 1:25people are shifting to conversations. In
47. 1:28fact, [snorts] Google just made the
48. 1:29biggest change in 25 years to their
49. 1:31search engine box. If you go to Google
50. 1:33right now and you type in something,
51. 1:35instead of it saying search, it says AI
52. 1:37mode. And when you click that button, it
53. 1:38actually goes to Gemini. It does not go
54. 1:40to traditional Google search.
55. 1:41>> So this part, he's also right, but the
56. 1:44shift underneath it started years ago.
57. 1:45Google used to rely much more heavily on
58. 1:48matching the words in your search to the
59. 1:49words on a page. Then in 2017,
60. 1:52researchers at Google published the
61. 1:54transformer paper. Transformer is the T
62. 1:56in chat GPT. It's basically the
63. 1:59architecture behind modern large
64. 2:01language models behind AI. Google then
65. 2:04started bringing that same generation of
66. 2:05language understanding technology into
67. 2:07search with systems like neural matching
68. 2:10and BERT. So, modern Google search and
69. 2:12chat GPT, they're not doing the exact
70. 2:14same thing, but they're built on a lot
71. 2:16of the same underlying ideas about
72. 2:18understanding language, context, and
73. 2:20meaning. This is why Google no longer
74. 2:22needs someone to type the exact same
75. 2:23words that appear on the page to rank
76. 2:26you.
77. 2:26>> We need to build out pricing content.
78. 2:28So, people are going to Gemini, Chatbt,
79. 2:31uh, Claude, and they're asking questions
80. 2:33and then they're saying, "Well, how much
81. 2:35does that cost to do?" And the reason
82. 2:36pricing content is so important is
83. 2:38because people ask AI pricing questions
84. 2:40constantly. And if you don't publish
85. 2:41pricing online, AI has nothing to
86. 2:43reference. And businesses that answer
87. 2:44this openly are going to win.
88. 2:46>> So I completely agree with this. We
89. 2:47actually built an entire content type
90. 2:49into the core 30 agent just for service
91. 2:51pricing guides. And the reason is very
92. 2:53simple. Local pricing content is still
93. 2:56massively underproduced. If someone in
94. 2:58Gary searches for water heater
95. 3:00replacement cost, and you're a plumber
96. 3:01in Gary, with a page that actually
97. 3:03explains what that service costs, what
98. 3:05makes the price go up or down, and what
99. 3:07somebody locally should expect to pay,
100. 3:08you've given Google and the AI systems a
101. 3:11really good source to site, and they
102. 3:13will almost always site a local source
103. 3:15if it's available. It's a huge
104. 3:17opportunity right now because most local
105. 3:18businesses still don't publish this. But
106. 3:21here inside the core 30 agent, we
107. 3:22actually have this whole section for
108. 3:24trust and you select which services
109. 3:26you're interested in. You fill out the
110. 3:28trust pages about the business and then
111. 3:30you can come in and fill out all the
112. 3:32information about the pricing itself.
113. 3:33Which means we can come here, we can
114. 3:35check the pricing guide and send this
115. 3:37over to the content queue and have this
116. 3:39written.
117. 3:39>> Imagine for home service businesses,
118. 3:41there's a native online booking tool,
119. 3:42whatever CRM that you use that allows
120. 3:45people to schedule right online, right
121. 3:46inside of these AI platforms. This is in
122. 3:49the next 6 to 12 months going to be a
123. 3:51very, very, very reality. I literally
124. 3:53was just talking to Jobber about this
125. 3:54exact thing. Yeah, this is the part I
126. 3:56think matters most for where local
127. 3:58search is going. We're absolutely
128. 4:00heading to a world where I can say find
129. 4:01me a plumber who can replace my water
130. 4:03heater Tuesday afternoon and the AI will
131. 4:06figure out who does that work, compare
132. 4:07the options and directly book the
133. 4:09appointment and like he said we can
134. 4:12already see major pieces of this
135. 4:13happening. Uh chat GPT can today show
136. 4:17restaurant availability through open
137. 4:18table, resi and Yelp and Open AI is
138. 4:20already testing quote requests for local
139. 4:22service businesses. We've also seen that
140. 4:24Google has an AI that can literally call
141. 4:26businesses and ask for openings or
142. 4:30availability. So this clearly is not
143. 4:32hypothetically. Now the booking itself
144. 4:34maybe it happens through Java or service
145. 4:35titan or whatever system that business
146. 4:37uses. But before the agent can book
147. 4:38anybody, it still has to decide who
148. 4:41belongs in that consideration set in the
149. 4:43first place. Who actually does water
150. 4:45heater replacement? Who serves this
151. 4:47area? Which business is trustworthy? who
152. 4:49has enough information available for the
153. 4:51agent to understand what they offer.
154. 4:54That's the part that I think becomes
155. 4:55really important later on in this
156. 4:57argument.
157. 4:58>> I used to be at a home service business.
158. 4:59The hardest thing was getting the
159. 5:00stories from my technicians, getting the
160. 5:02stories from my CSRs, and actually
161. 5:04getting real stuff. We had to try to
162. 5:06come up with it our own. So, what people
163. 5:09need to be doing is every phone call
164. 5:12your company receives contains future
165. 5:14content. Document questions, answer them
166. 5:16publicly. this is how you train AI to
167. 5:18trust you.
168. 5:19>> I mean, again, I largely agree with what
169. 5:21he's saying, but I would go a little bit
170. 5:23farther than just using real customer
171. 5:24questions. Now, we do a ton of research
172. 5:27and data calls inside the core 30 agent
173. 5:29before we write anything because the
174. 5:30goal is to make the pageformationally
175. 5:32additive, right? We want search console
176. 5:35data, competitor research, reviews, real
177. 5:37customer questions, local information.
178. 5:39We'll even make calls up to services
179. 5:41like the US Census Bureau when the page
180. 5:43needs local demographic or geographical
181. 5:45context. Because if all you're doing is
182. 5:47taking the same generic answer and
183. 5:48rewriting it with AI, I don't think that
184. 5:51gets better just because the wording
185. 5:52came from a customer call. The useful
186. 5:54part is adding information that's
187. 5:56actually specific to that business, to
188. 5:59that service or to that market.
189. 6:02>> Agents talking to agents. So imagine you
190. 6:06order something. Where's my package?
191. 6:08You're and and you ordered it from
192. 6:09Target. Your personal agent, let's just
193. 6:11say you use Gemini Spark. If you know,
194. 6:13you're going to get enrolled into that
195. 6:14anyways, probably. Your agent goes to
196. 6:16Target and is interacting with Target's
197. 6:18agent. Okay. Then the Target agent says,
198. 6:22"Okay, well, let me go check on UPS's
199. 6:24system." Okay. Now, there's three agents
200. 6:26involved. There's Gemini Spark, there's
201. 6:28Target's agent, and there's UPS's agent.
202. 6:30Where we stand right now, pe these
203. 6:32companies are spending millions of
204. 6:33dollars investing in these agents, and
205. 6:36there's no protocol to connect all of
206. 6:37them. What Google rolled out is the
207. 6:39protocol to connect all of them so that
208. 6:41they can interact and come back to you
209. 6:42with an answer. So you have something
210. 6:45that you want, you reach out to your
211. 6:46agent, your agent goes and figures it
212. 6:48out and then comes back with an answer
213. 6:49instantly. No search, no website, no
214. 6:51calls, nothing.
215. 6:52>> So keep in mind in this example, you
216. 6:55already know that you that the
217. 6:57information lives with target, right?
218. 7:00Queries target system. You said, "Hey, I
219. 7:02did something with Target. Please go
220. 7:03find this answer at Target." the agent
221. 7:06isn't trying to figure out where to go
222. 7:08to find the answer. It already knows
223. 7:10where to go. I think that's an important
224. 7:12consideration here uh based on the leap
225. 7:16of logic that we come to.
226. 7:17>> And that is because
227. 7:20Target has an agent. So, are people
228. 7:22going to go to Target to buy things or
229. 7:23are they going to go to their agent and
230. 7:24say, "Hey, can you buy something for
231. 7:25me?" Go to Target and figure it out.
232. 7:28Imagine local businesses, home service
233. 7:29businesses. when when people go and ask
234. 7:33their agent, hey, can you go and do
235. 7:35information? Can you go find things for
236. 7:36me? What's going to happen is we're in
237. 7:39the crawl era right now. It's crawling
238. 7:40websites, which is what I just showed
239. 7:41you. You know, the pricing page, the
240. 7:43content, uh the online scheduling. But
241. 7:46what if there was a place where you had
242. 7:48all of your knowledge, all of your
243. 7:50information, not on a website, but in
244. 7:52what Google calls a knowledge catalog?
245. 7:54>> So, again, I agree with the destination.
246. 7:56We're definitely heading toward a world
247. 7:57where your AI assistant can do a lot of
248. 7:59the searching and comparing for you and
249. 8:01eventually interact directly with the
250. 8:02systems the businesses use. And as you
251. 8:04mentioned, Google is building that
252. 8:06infrastructure for agents to talk to
253. 8:08other agents. OpenAI is moving toward
254. 8:10actions inside ChatGpt. So, it's not
255. 8:12some crazy science fiction version of
256. 8:15search. But there's a question
257. 8:17underneath all of this that matters a
258. 8:19lot for an actual local business, right?
259. 8:22Before the agent can contact anybody,
260. 8:25who does it choose? If I ask for a
261. 8:27plumber to replace a water heater, there
262. 8:29might be 50 plumbers in that market, the
263. 8:31agent still has to figure out which ones
264. 8:33actually do that service, which ones
265. 8:35serve my area, and which ones it trusts
266. 8:37enough to put in front of me. That
267. 8:39decision, the data to make that decision
268. 8:41has to come from somewhere. And a
269. 8:44knowledge catalog is a verified data set
270. 8:46of information about your business where
271. 8:49you can then deploy it as an agent on
272. 8:50your behalf to interact with it. This
273. 8:52might sound all woo woo and crazy,
274. 8:54whatever, but Google's literally talking
275. 8:56about this for businesses. Like you got
276. 8:58to have a knowledge catalog. They
277. 8:59already rolled that out. You got to have
278. 9:01uh the connection with Google Search
279. 9:03Console with the verified domain. They
280. 9:04already rolled that out. You have to
281. 9:05have a discoverability API. They rolled
282. 9:07that out. They're rolling out all these
283. 9:09things to give people the assets, but
284. 9:10most businesses don't know about this.
285. 9:12So, this is where we're starting to
286. 9:13split, right? I agree that the customer
287. 9:15is not likely going to visit your
288. 9:17website in the future. I agree the agent
289. 9:19is eventually going to be able to do the
290. 9:21comparison, make the call, or book the
291. 9:22appointment, but that doesn't mean the
292. 9:24website stops mattering. The agent still
293. 9:26needs the information to decide which
294. 9:28business deserves the recommendations in
295. 9:29the first place. For a local business,
296. 9:31the Google business profile is still the
297. 9:33center, the heart of that local business
298. 9:35entity. The website gives much more
299. 9:38depth around what services the business
300. 9:40actually provides, where it provides
301. 9:42them, what those services cost, and
302. 9:43whether they're relevant to what the
303. 9:44customer is asking for. And then the
304. 9:46rest of the internet gives the agent
305. 9:48cooperation, reviews, citations, links,
306. 9:51local mentions, other sites talking
307. 9:52about that same business. So the
308. 9:54interface can absolutely change, but I
309. 9:57don't think removing the customer from
310. 9:59the website removes the website from the
311. 10:01decision. So the distinction he's trying
312. 10:04to make here, he's separating the the
313. 10:06website that people see from the
314. 10:09underlying business information an agent
315. 10:11would actually need. Services, service
316. 10:13areas, pricing, policies, credentials,
317. 10:15FAQs, all of that. And that part makes
318. 10:17sense. Google has something called a
319. 10:19knowledge catalog. That's what he's
320. 10:21talking about. That exists. But right
321. 10:23now, it's an enterprise infrastructure.
322. 10:25It's not like a Google business profile
323. 10:27where a plumber can log in and fill out
324. 10:29all this information. Now, yes, Google
325. 10:31could eventually expand the business
326. 10:32profile into something richer and more
327. 10:35machine readable, but you know,
328. 10:38absolutely they could, and it wouldn't
329. 10:39surprise me if they did at all. But even
330. 10:41if they went in that direction, I don't
331. 10:43think that replaces the website. A
332. 10:46structured catalog can tell an agent
333. 10:47that you replace garbage disposals, but
334. 10:50right, your website can tell it that you
335. 10:52replaced one at 7:00 last night for a
336. 10:54homeowner whose sink backed up during
337. 10:56dinner. Those are not the same kind of
338. 10:58information. One is structured business
339. 11:00data. The other is evidence, context,
340. 11:03and experience. And agents are going to
341. 11:05need both. That's one of the reasons
342. 11:07that I like the core 30 structure so
343. 11:09much. Instead of one generic services
344. 11:12page with 15 different things jammed
345. 11:13into it, every service entity gets its
346. 11:16own page with a clear relationship back
347. 11:18to the business. So if an agent is
348. 11:20trying to understand whether this
349. 11:21plumber actually does water heater
350. 11:23replacement or sewer repair or garbage
351. 11:26disposal replacement, there's a very
352. 11:28clean source of information for it to
353. 11:30work from. And that is good local SEO
354. 11:33today. It also it also happens to make a
355. 11:36lot of sense for the agentic version of
356. 11:38search that he's describing.
357. 11:40>> So let me just go back and forth. Old
358. 11:42internet, new internet. What drives
359. 11:44traffic? The old approach is SEO,
360. 11:46keywords, and backlinks. The new
361. 11:47approach is AI trusted content and data.
362. 11:49This is why you need to create the real
363. 11:51content with real customer questions
364. 11:52because it's not just about jamming a
365. 11:54bunch of keywords anymore. It's about
366. 11:55conversational AI search. How customers
367. 11:58find you? Before they did search,
368. 11:59website lead. Now it's going to be AI
369. 12:02answers. Then the agent takes action on
370. 12:04both parties behalf and takes care of
371. 12:06this. It takes care of it. What your
372. 12:08content is for rankings and engagement.
373. 12:11>> Okay, so he just said rankings aren't
374. 12:12going to matter. This is another place
375. 12:14where I think he just goes too far.
376. 12:16Rankings will absolutely still matter in
377. 12:18an AI agent world. An AI agent is not
378. 12:21going to spin the compute to deeply
379. 12:23evaluate every single plumber in a city
380. 12:26every single time somebody asks for a
381. 12:28recommendation that has to do with
382. 12:30plumbing. It must narrow the field
383. 12:32first. And that's where ranking and
384. 12:34retrieval still matter. If your business
385. 12:36makes that first candidate set, then the
386. 12:38AI can evaluate whether you're the best
387. 12:40fit for that particular person or not.
388. 12:43But maybe you're too far away. Maybe you
389. 12:45don't offer that exact service. Maybe
390. 12:46your pricing and availability don't fit.
391. 12:48That's fine. But if you're the 50th
392. 12:51highest plumber in that city, there's a
393. 12:53really good chance the system never
394. 12:55spends the compute to evaluate your
395. 12:57business for that consumer in the first
396. 12:59place. And the studies we have so far
397. 13:01point in that direction. One study of
398. 13:04chat GPT citations found that the top
399. 13:06retrieval result was cited 58% of the
400. 13:09time. by position 10, still on the first
401. 13:11page, but position 10 that had fallen to
402. 13:1414%. Another study of 400,000 Google AI
403. 13:17overviews found that 52% of the sources
404. 13:20they cited also ranked in the top 10.
405. 13:22And this gets even more interesting in
406. 13:24local. Bright Local analyzed 200,000
407. 13:27local AI searches and found that the
408. 13:29systems typically recommend only two to
409. 13:31four businesses. They also found that
410. 13:33business websites accounted for 42% of
411. 13:36the citations they analyzed. So all of
412. 13:38this does not mean that rankings
413. 13:40disappear. Ranking increasingly
414. 13:42determines whether you make the set that
415. 13:44the AI evaluates at all. And I'd also
416. 13:47want to be careful with the phrase train
417. 13:49AI to trust you. Publishing a page on
418. 13:52your website is not the same thing as
419. 13:53training chat GPT or Gemini. What you're
420. 13:56really doing is putting better
421. 13:57information online for these systems to
422. 13:59retrieve. They can then compare it
423. 14:01against your business profile, your
424. 14:02reviews, your citations, other websites,
425. 14:04whatever else they have available. And
426. 14:06if all of that lines up, you've made it
427. 14:08easier for the system to understand what
428. 14:10your business does and decide whether
429. 14:11you're relevant enough to surface. It's
430. 14:14different from training AI. And it means
431. 14:16that your website and rankings are still
432. 14:18important, right? You still need to get
433. 14:21in the door. You still need to do enough
434. 14:23to be considered.
435. 14:24>> Way less than that, that you can
436. 14:25actually deploy these things in your
437. 14:26business and get an advantage. Build a
438. 14:29knowledge catalog, FAQs, services,
439. 14:31policies, service areas, expertise. I
440. 14:33can't stress this one enough. You got to
441. 14:34be documenting all of your stuff.
442. 14:36Optimize for AI citations. Focus on
443. 14:38being recommended, not just ranked. So,
444. 14:40right now, people will go to chatbt and
445. 14:42say, uh, you know, who's the best
446. 14:44plumber near me or whatever. And then
447. 14:45they'll evaluate if they're showing up
448. 14:46for that keyword or not. Um, I made a
449. 14:48whole video on this, which is how to
450. 14:50actually track your rankings in AI
451. 14:52search. And it's to shift from keywords
452. 14:54like what you are traditionally used to
453. 14:56doing to prompting and actual like
454. 14:58tracking prompts. And there's two
455. 15:00things, share a voice and mention rate,
456. 15:01which are the most important ones. And
457. 15:03when I say share a voice, it's how many
458. 15:05times that they share your information,
459. 15:07which is why you can't just rely on a
460. 15:08blog post. You have to have multiple
461. 15:09different types of content to be shown.
462. 15:11And think of every piece of content as
463. 15:13training data for the future internet.
464. 15:14>> So, I understand what he's getting at
465. 15:16here, but I think training data is not
466. 15:18the best way to describe this. The
467. 15:20opportunity is not that every page you
468. 15:22publish somehow teaches the model who
469. 15:24you are. It's that every useful page
470. 15:26gives these systems another piece of
471. 15:28evidence that they can retrieve when
472. 15:30they're trying to understand your
473. 15:31business a little bit better. And that
474. 15:33brings me back to where I disagree with
475. 15:34the title. If the agents are going to do
476. 15:37more of the searching, comparing, and
477. 15:38eventually booking for us, that doesn't
478. 15:40make the website less important. It does
479. 15:43change what a good website needs to do.
480. 15:45Historically, we've built websites for
481. 15:47human users and optimized them for
482. 15:49Google's algorithm. But now the website
483. 15:51needs to make the business incredibly
484. 15:53easy for a machine to understand. What
485. 15:56do you do? Where do you do it? What does
486. 15:58it cost? What have you actually done
487. 15:59before? And then the rest of the
488. 16:01internet has to cooperate that you're a
489. 16:03real business worth recommending. And
490. 16:06agreed the customer probably will never
491. 16:08visit your website, but the agent still
492. 16:10will. So, if you actually want to see
493. 16:12what this looks like in practice, I have
494. 16:14a video where I break down the exact
495. 16:16site structure I use for local
496. 16:17businesses so Google and AI systems can
497. 16:19understand every service they offer.
498. 16:21That's the one that I would watch