Skip to content
ChatGPT · Google · OpenAIEpisode 017 · 4 August 2024 · 36:01

ChatGPT Is Changing Search, but the Web Is Not Disappearing—What Changes Is Who Captures the Value

Central question

If ChatGPT changes search, who captures the value when the web remains but user behavior changes?

What you take away

Identify who captures value when ChatGPT and OpenAI change search and the user journey. A practical assessment requires the reader to track who controls the source of the answer, traffic, data, and the user’s next choice.

Main threads

What to watch for

1Compare “ChatGPT replaces search?” with “What happens to SEO when keyword-stuffed sites die?”: they provide different criteria for judging the same issue.
2Test the conclusion from “A minute of pessimism from Alexander Mashrabov: will generative AI kill nothing and replace nothing?” in your own use case—what actually changes in the process and what remains a promise.
3Before choosing a product or approach, record the constraint identified in “Facebook (META), the Llama model, ChatGPT retention, and Zuckerberg's ambitions”.
4Define the owner of the outcome and the quality metric for the situation described in “X/Twitter training on user posts”.
Signals to track afterwards
Watch for actions by TechCrunch and Amazon that confirm or challenge the episode’s central claims.
Compare new launches and policy changes with “A minute of pessimism from Alexander Mashrabov: will generative AI kill nothing and replace nothing?”: have access, quality, price, or constraints changed?
Check whether the scenario in “X/Twitter training on user posts” becomes repeatable practice rather than a one-off demonstration.
Most useful for
Product teamsMedia professionalsMarketersAI usersExecutives and managersInternet users

Key takeaways

00:56ChatGPT is already convenient as a search tool: formulate a complex question, refine the parameters, and receive a synthesized answer

The “ChatGPT replaces search?” issue should be assessed with one constraint in mind: ChatGPT is already convenient as search — a complex question, refined parameters, a synthesized answer — and the main change is not in the technology but in expectations: people no longer want to open ten pages to assemble a conclusion themselves.

03:36The market tests it through use: how ChatGPT falls short of a search engine. On hallucinations

For the “How ChatGPT falls short of a search engine. On hallucinations” scene, the decisive point is this: the model can confidently invent nonexistent books or confuse a source, so it does not yet replace a search engine — it changes user expectations while leaving fact-checking on the user's side.

09:30For SEO, this is a painful shift

The boundary of the “What happens to SEO when keyword-stuffed sites die?” case is defined by this point: sites stuffed with keywords and created only to capture a click from Google lose their purpose. Useful material is still needed, but the interface between it and the reader may belong not to the publisher, but to the model. If ChatGPT summarizes a page, the site paid to produce the content while another product received the relationship with the audience.

12:54Who owns the outcome: how websites fight OpenAI

The practical meaning of “How websites fight OpenAI” is that TechCrunch and other outlets restrict model access while Forbes and Condé Nast press claims against AI search services, but a total ban is dangerous: if one source will not open, the user simply moves to another — publishers need a model where citation, licensing, and traffic make economic sense again.

14:50What changes in real work: X/Twitter training on user posts

The “X/Twitter training on user posts” topic becomes clearer once this point is included: X goes against the blocking trend and trains on user posts: data created every second turns into a model advantage with almost no separate distribution effort.

15:39That is why access blocks and lawsuits are appearing

The “A minute of pessimism from Alexander Mashrabov: will generative AI kill nothing and replace nothing” topic becomes clearer once this point is included: search is a "love triangle" of advertiser, platform, and user that Google once solved with its auction; generative search improves things only for the user, unit economics break on expensive LLM calls, and none of the hosts has ever seen anyone make a purchase decision through Perplexity.

21:44How the issue moves from news to product: does the AI economy add up? How many years

The boundary of the “Does the AI economy add up? How many years was Amazon unprofitable?” case is defined by this point: OpenAI loses billions, video generation demands enormous spending, and user retention is still unproven; Amazon was unprofitable for a long time too — but that does not make every loss justified.

24:11X is moving in the opposite direction and using user posts for training

The “Facebook (META), the Llama model, ChatGPT retention, and Zuckerberg's ambitions” topic becomes clearer once this point is included: meta is integrating Llama into WhatsApp and Facebook and wants to overtake ChatGPT in audience size. These companies possess data created every second and can convert it into a model advantage with almost no separate distribution effort.

What this episode is about

People increasingly ask ChatGPT instead of Google, websites block model access, and publishers sue AI search companies. The threat to web publishers is not that pages will vanish, but that they will lose the direct relationship with the user—and the money that used to arrive with the click.

ChatGPT is already convenient as a search tool: formulate a complex question, refine the parameters, and receive a synthesized answer. But the model can confidently invent nonexistent books or confuse a source. It does not yet replace a search engine. It changes what users expect. People no longer want to open ten pages and assemble the conclusion themselves.

For SEO, this is a painful shift. Sites stuffed with keywords and created only to capture a click from Google lose their purpose. Useful material is still needed, but the interface between it and the reader may belong not to the publisher, but to the model. If ChatGPT summarizes a page, the site paid to produce the content while another product received the relationship with the audience.

That is why access blocks and lawsuits are appearing. TechCrunch and other outlets restrict access; Forbes and Condé Nast raise claims against AI search services. But a total ban is dangerous too: if one source cannot be opened, the user may simply move to another. Publishers will have to find a model in which citation, licensing, and traffic once again make economic sense.

X is moving in the opposite direction and using user posts for training. Meta is integrating Llama into WhatsApp and Facebook and wants to overtake ChatGPT in audience size. These companies possess data created every second and can convert it into a model advantage with almost no separate distribution effort.

The economics of AI still do not add up perfectly: OpenAI loses billions, video generation requires enormous spending, and user retention remains to be proven. Amazon was unprofitable for a long time too, but that does not make every loss justified.

The winner will be the system that not only answers better, but also creates a sustainable bargain among the user, the platform, and the people who produced the source information.

A convenient answer does not eliminate the web; it changes who receives attention, money, and the right to choose sources on the user’s behalf.

Episode transcript

The episode is in Russian; below is an English reading guide to the transcript (the full EN transcript is a machine translation). Voice matching applied to 70 segments: 38 identified, 3 mixed, 16 probable, and 13 unresolved.

Loading…