Technologist Mag
  • Home
  • Tech News
  • AI
  • Apps
  • Gadgets
  • Gaming
  • Guides
  • Laptops
  • Mobiles
  • Wearables
  • More
    • Web Stories
    • Trending
    • Press Release

Subscribe to Updates

Get the latest tech news and updates directly to your inbox.

What's On
The Best Fitness Trackers of 2026: Garmin, Google Fitbit, and More

The Best Fitness Trackers of 2026: Garmin, Google Fitbit, and More

19 June 2026
This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

19 June 2026
Shinobi: Art of Vengeance Jumps To Switch 2 In September

Shinobi: Art of Vengeance Jumps To Switch 2 In September

19 June 2026
Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

19 June 2026
AI agents need more than reasoning: they need to actually use the web

AI agents need more than reasoning: they need to actually use the web

19 June 2026
Facebook X (Twitter) Instagram
Facebook X (Twitter) Instagram
Technologist Mag
SUBSCRIBE
  • Home
  • Tech News
  • AI
  • Apps
  • Gadgets
  • Gaming
  • Guides
  • Laptops
  • Mobiles
  • Wearables
  • More
    • Web Stories
    • Trending
    • Press Release
Technologist Mag
Home » AI agents need more than reasoning: they need to actually use the web
Tech News

AI agents need more than reasoning: they need to actually use the web

By technologistmag.com19 June 20268 Mins Read
AI agents need more than reasoning: they need to actually use the web
Share
Facebook Twitter Reddit Telegram Pinterest Email

A company rolls out an AI customer service assistant. The model behind it is current and capable enough for the job. The assistant goes live. Within a week, support tickets are getting worse, not better.

The model isn’t the problem. The company’s own website is. The return policy the assistant needs to quote lives in a PDF. The shipping calculator it needs to reference is a multi-step form. The product specs it should be pulling sit behind a tabbed interface that only loads after a click. To a human visitor, the site works fine. To the AI trying to read it, half the site doesn’t exist.

This is the wall most agentic AI deployments are hitting right now, and it has almost nothing to do with the model.

McKinsey’s 2025 State of AI report found that 23% of organizations are now scaling agentic AI systems in at least one business function, with another 39% experimenting. Most of those deployments will run into the same wall: a web designed for humans, used by software that needs something humans never required. The next step for AI agents isn’t smarter reasoning. It’s the ability to actually navigate and use the live internet.

The three things an AI agent has to do on the web

The work breaks down into three jobs, and all three have to work for an agent to be useful in production.

Search. The agent needs to find the right information. Not URLs to a list of links, but actual content it can read and reason over. If a customer asks an insurance chatbot whether their policy covers a specific event, the agent needs to surface the relevant section of the policy, not a search results page.

Scrape. Once the agent finds the page, it needs to read it cleanly. Most modern websites don’t make this easy. Pages load through JavaScript that has to execute first. Content lives inside expandable accordions, tabs, and lazy-loaded sections. The HTML the agent receives often looks nothing like what a human sees in their browser.

Interact. This is where most agent demos fall apart in production. A lot of the information humans care about isn’t on a simple URL. It’s behind a “load more” button, a search box, a multi-step form, a navigation menu, or a login. A scraper that can only read static pages can’t reach any of it. An agent that can interact (click, navigate, fill, submit) can. The difference between the two determines whether the AI can actually do its job.

Of the three, interaction is the newest and the hardest. It’s also where the most useful agent applications live: shopping assistants that compare prices across sites, research tools that pull data from interactive dashboards, customer support bots that navigate documentation portals the way a real user would.

Firecrawl is building the layer underneath

Firecrawl is one of the companies building infrastructure designed to support all three functions. The platform sits between AI agents and the live web, handling search, scraping, and interaction as managed capabilities behind a single API. Its open-source project has more than 120,000 stars on GitHub. Customers including Lovable, Replit, and Zapier  use it in production. Nexus Venture Partners led the company’s $14.5 million Series A in 2025, with Shopify CEO Tobi Lütke joining as an investor after first using Firecrawl as a customer.

The pitch is straightforward: an AI agent built on top of Firecrawl doesn’t need its development team to write custom code for every site it touches. It calls an API, and the platform handles much of the underlying technical work: rendering JavaScript, navigating dynamic pages, interacting with elements, and returning structured output that AI systems can use.

“Every AI company needed clean web data and nobody was solving it well,” says Eric Ciarla, one of Firecrawl’s cofounders. “So we built Firecrawl.”

Ciarla and his cofounders ran into the problem directly while building their previous company, Mendable, an AI search platform that was used by a range of organizations. The search product worked. The infrastructure pulling data from each customer’s website to feed it didn’t. Every new integration meant rebuilding fragile extraction code that broke the next time the customer’s site changed. Mendable wasn’t unusual in hitting that wall. Many AI companies integrating web data faced similar challenges, repeatedly rebuilding internal extraction tools.

How AI is becoming the new way people find things

There’s a shift happening alongside the technical one, and it changes the stakes for businesses that haven’t thought about AI agents reading their websites yet.

For two decades, the path from “a customer is looking for something” to “a customer finds your business” often ran through traditional search engines. AI assistants are increasingly where people start when they want a recommendation, a comparison, or an answer. The AI assistant goes off, pulls information from the relevant websites on the person’s behalf, and comes back with a synthesized answer. If the AI couldn’t parse your site, your business doesn’t appear in the answer.

Ciarla argues this changes how businesses should think about AI crawlers entirely. “Behind every AI agent is a human trying to find something,” he says. The dominant industry framing has treated AI crawlers as unwelcome automation: bots to defend against, traffic that drains server resources without sending human visitors in return. That framing made sense when the only things reading websites at scale were search engines indexing for human visitors later. It makes less sense when AI agents are the path the human is using to find.

 In Ciarla’s view, blocking AI crawlers today may be comparable to limiting visibility on an emerging discovery channel. He argues that doing so could reduce opportunities for businesses to be found through evolving customer search behaviors.

What makes Firecrawl’s position in this shift unusual is that it doesn’t require businesses to do anything. Most approaches to AI visibility put the work on the site owner: add new markup, expose new endpoints, restructure pages, learn a new optimization discipline on top of the existing SEO one. Firecrawl works from the opposite direction. The platform handles the conversion between human-readable site and machine-readable data automatically, in real time. A business never needs to know AI agents are reading the page. The agents get what they need anyway.

The bigger question underneath

As agents pull more information from more sites, the relationship between AI systems and the sources they depend on becomes a real question. A model where AI extracts value from web content without anything flowing back to the people who created it isn’t durable. Publishers are pushing back through lawsuits and access blocks, and major sites are increasingly walling off their content from AI crawlers entirely. The underlying ecosystem isn’t healthy, and the long-term cost lands somewhere eventually.

In March 2026, Firecrawl partnered with Wikimedia Enterprise to route all of its Wikipedia traffic – 2 to 3 million requests per month – through Wikimedia’s commercial APIs rather than continuing to scrape Wikipedia pages directly. The arrangement replaces resource-intensive scraping with paid, structured access, and helps support the volunteer community that maintains one of the most-cited information sources on the open web.

“The community members who write and edit these articles hold immense power in the age of AI,” Ciarla said when the partnership was announced. “They are providing the essential service of defining what is true. We want to ensure our infrastructure supports their work rather than just consuming it.”

The Wikimedia deal is one model. Similar approaches may emerge elsewhere in the industry. As AI products move from demos into production at scale, the companies building the underlying infrastructure are helping shape how AI systems interact with the web.

What this means if you’re paying attention

If you’re building with AI, the practical takeaway is simple. The model is no longer the differentiator. Almost everyone has access to the same frontier models, and the gaps between them keep closing. What separates an AI product that works in production from one that doesn’t is increasingly the layer underneath, and whether the system can actually reach the information it needs to be useful. Investing in that layer may offer meaningful engineering benefits.

If you’re running a business and you’ve never thought about AI agents reading your website, that’s the moment to start. The discovery channel is shifting. A customer who previously may have found a business through a traditional search engine may now use an AI assistant as part of the discovery process. If that assistant can’t read your site, they may not find you at all. Many businesses continue to optimize primarily for human readers and search engines while evaluating how AI-driven discovery may affect their digital presence.

Digital Trends partners with external contributors. All contributor content is reviewed by the Digital Trends editorial staff.

Share. Facebook Twitter Pinterest LinkedIn Telegram Reddit Email
Previous ArticleThe Most Promising Ebola Vaccine Has Been Sitting on the Shelf for 15 Years
Next Article Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

Related Articles

The Best Fitness Trackers of 2026: Garmin, Google Fitbit, and More

The Best Fitness Trackers of 2026: Garmin, Google Fitbit, and More

19 June 2026
This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

19 June 2026
Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

19 June 2026
The Most Promising Ebola Vaccine Has Been Sitting on the Shelf for 15 Years

The Most Promising Ebola Vaccine Has Been Sitting on the Shelf for 15 Years

19 June 2026
This monstrous ASUS gaming laptop costs as much as three new MacBook Pros

This monstrous ASUS gaming laptop costs as much as three new MacBook Pros

19 June 2026
The Best Art TVs | WIRED

The Best Art TVs | WIRED

19 June 2026
Stay In Touch
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Vimeo

Subscribe to Updates

Get the latest tech news and updates directly to your inbox.

Don't Miss
This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

This free app gives your photos the Game Boy Camera’s iconic look, no cartridge needed

By technologistmag.com19 June 2026

The blocky, low-res photo that once lived only on your Game Boy screen can now…

Shinobi: Art of Vengeance Jumps To Switch 2 In September

Shinobi: Art of Vengeance Jumps To Switch 2 In September

19 June 2026
Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

Pump.Fun’s Bounties Platform Is a Black Hole of Circular Grifting

19 June 2026
AI agents need more than reasoning: they need to actually use the web

AI agents need more than reasoning: they need to actually use the web

19 June 2026
The Most Promising Ebola Vaccine Has Been Sitting on the Shelf for 15 Years

The Most Promising Ebola Vaccine Has Been Sitting on the Shelf for 15 Years

19 June 2026
Technologist Mag
Facebook X (Twitter) Instagram Pinterest
  • Privacy
  • Terms
  • Advertise
  • Contact
© 2026 Technologist Mag. All Rights Reserved.

Type above and press Enter to search. Press Esc to cancel.