Rethinking SEO in the Age of AI: Key Takeaways from Eli Schwartz
Updated: 6 days ago
In this episode of Lenny's Podcast, SEO advisor and Product-Led SEO author Eli Schwartz explains why artificial intelligence is changing search without making SEO irrelevant. The discussion moves beyond rankings and traffic to a more important question: does search fit the way a company's customers actually discover, evaluate, and buy its product?
Schwartz draws on work with companies including Coinbase, Tinder, LinkedIn, WordPress, Zapier, and Quora. His central argument is that SEO should be treated as a product and business decision—not as a publishing contest measured by how many pages a company can produce.
AI Overviews Are Changing the Search Results Page
The conversation begins with Google's response to AI chatbots. Search Generative Experience, later renamed AI Overviews, placed generated answers directly inside search results. Its early development raised difficult questions about monetization, attribution, liability, and the risk of reproducing material from publishers.
Despite those complications, AI-generated answers continued to expand. That changes the visual and economic structure of a search results page: a user may receive a broad answer before reaching the familiar list of organic links.
Schwartz does not interpret this as the end of search. Instead, he sees AI as particularly effective during the early discovery stage. A generated overview can give someone a starting point, while deeper evaluation may still require traditional results, specialist sources, product pages, reviews, or direct action.
Search Intent Matters More Than Content Volume
AI search is strongest when the user wants a general explanation. This creates pressure on websites whose value consists mainly of reorganizing widely available information. Broad top-of-funnel articles and generic software lists are easier for a search engine to condense into a direct answer.
Pages serving a specific need later in the customer journey are harder to replace. If a website helps a user compare a real decision, complete a task, understand a specialized situation, or use a product, it offers more than a generic overview.
This is why Schwartz emphasizes intent. The right question is not simply, “Which keyword has the most searches?” It is, “What is this person trying to accomplish, and can our page genuinely help?”
Product-Led SEO Starts With the User's Problem
Two examples show how SEO can be connected to a product rather than placed beside it as a separate content operation.
Zapier recognized that users wanted different software products to work together. Pages about connecting tools such as Gmail and Salesforce mapped directly to that need and demonstrated what Zapier could do. The search page was therefore part of the product experience.
Tinder approached location-based searches through the underlying problem of meeting people in a new city. Instead of treating local pages as arbitrary keyword targets, the strategy connected a search situation to the reason someone might use Tinder.
Both cases begin with a user problem and work backward toward a searchable asset. Neither begins with a request to publish a large number of articles.
A Three-Step Way to Start SEO
Schwartz describes a practical sequence for deciding what to build.
First, understand the user. Identify what they need, how they describe the problem, and what they would realistically search for. Internal product language is often different from the language customers use.
Second, decide what kind of asset can satisfy that need. It may be an editorial guide, a landing page, a template, a comparison, a database, or a programmatically generated page. The format should follow the use case rather than a fixed content calendar.
Third, build for the search user as if the page were a product. Design, engineering, user research, information structure, and usability can all matter. A page that ranks but fails to help the visitor is not a successful SEO product.
When SEO Is the Wrong Growth Channel
One of the video's strongest points is that not every company needs SEO. Some products solve a problem users do not yet know they have, so there is little existing search demand. Other products require long enterprise sales processes involving many stakeholders, making search only a small part of the journey.
Local businesses may also be better served by platforms where customers already make decisions. A restaurant, for example, may receive more value from Google Maps and delivery marketplaces than from a large editorial website.
Schwartz frames growth around four broad engines: SEO, paid acquisition, virality, and sales. An early-stage company with limited resources should compare them rather than assume that organic traffic is automatically the cheapest choice.
SEO Is an Investment, Not Free Traffic
SEO consumes time, people, engineering work, editorial effort, and sometimes substantial direct expense. Its expected return should be compared with influencer partnerships, brand advertising, paid campaigns, outbound sales, and other available channels.
The correct success metric depends on the business. A media company may care about page views and repeat readership. A SaaS company should pay more attention to qualified leads, product use, or revenue. Traffic that never produces a meaningful business outcome can create the appearance of success while consuming resources.
SEO also works on different timelines for different companies. An established brand may see results sooner because it already has authority and demand. A new company may need a longer period to earn visibility. Schwartz recommends treating SEO like product development, with milestones, hypotheses, and a defined point at which the company evaluates whether the investment is working.
AI Can Produce Content, but It Cannot Supply the Strategy
The video takes a pragmatic view of AI-generated content. AI can be useful for bounded production tasks, such as creating product descriptions across a large ecommerce catalog. In that situation, the underlying value comes from the products and structured data, while the text helps explain them.
AI becomes less useful when it produces large amounts of interchangeable material without a clear user need. As publishing becomes cheaper, the internet fills with more content, which makes usefulness and differentiation more—not less—important.
The strategic work still requires choosing the right user, problem, format, and business objective. A language model can help create a page, but it cannot make an irrelevant page valuable simply by making it longer.
AI Overviews May Be More Valuable for Branding Than Clicks
Schwartz suggests that appearing in an AI-generated answer can function as brand exposure. A prominent brand mention signals familiarity and authority, even if the user does not immediately visit the site.
The value of a citation link is less certain because search engines can use a publisher's information while reducing the need to click. This means companies should consider visibility and brand recognition alongside traditional ranking and traffic measures.
That distinction is also useful when comparing AI SEO agencies: ask how they report brand citations and recognition alongside qualified visits, leads, and revenue.
Brand Building and SEO Should Reinforce Each Other
The conversation rejects the idea that link building is an isolated technical exercise. Strong brands earn mentions because people know them, trust them, and consider their work relevant. Links gained through genuine authority and useful content are more meaningful than unrelated placements acquired only to influence rankings.
SEO should therefore support the company's product and brand rather than operate as a detached traffic project. Content that attracts an audience with no relationship to what the company offers may increase analytics numbers without improving the business.
Programmatic SEO Versus Editorial SEO
Programmatic SEO works when a reliable data source can generate a useful page for a repeated user need. TripAdvisor can combine hotel, location, review, and traveler data. Zillow can combine property details, photos, pricing, and neighborhood information. The scale comes from the underlying dataset, while each page answers a distinct query.
Editorial SEO requires someone to research and write each page. Producing a handcrafted article for every hotel or property would be prohibitively expensive and might not improve the user experience.
The choice is therefore not “automation good” or “automation bad.” Programmatic pages make sense when structured data and repeated intent create genuine utility. Editorial work makes sense when judgment, explanation, reporting, or a distinctive point of view is necessary.
SEO Will Survive Because People Still Want Choice
The final sections address the idea that answer engines will replace search entirely. Schwartz argues that users often want more than one answer, particularly when money, taste, risk, or personal preference is involved. Search lets them compare sources and choose what to trust.
AI may reduce some informational searches, but people will continue to investigate products, companies, services, and competing perspectives before acting. SEO will change as the interface changes, yet the need to be discoverable during those decisions remains.
The Main Takeaway
The video's message is not that companies should abandon SEO. It is that SEO must earn its place in the growth strategy. A successful program begins with user intent, produces an asset suited to that intent, connects naturally to the product, and measures outcomes that matter to the business.
AI makes generic answers easier to generate and harder to monetize. That raises the value of useful data, clear brand positioning, specialized knowledge, and pages that help users do something. Companies that treat SEO as a product discipline will be better prepared than those that treat it as a race to publish more words.
If you regularly collect long interviews, market research, campaign notes, and strategic documents, remio can help keep those materials searchable in one personal knowledge base. You can also download remio to explore it directly.
