OK, so in this article I’m going to go into a topic that I haven’t seen a lot of others writing about: Agentic SEO.
Recently, I was having a conversation around AI search and the person I was talking to asked me what I thought was next. They wanted to really skate where the puck was going and have conversations about what things could look like in 2-3 years.
My immediate reaction was to tell them that they need to be considering Agentic SEO. In a world where users are deploying AI agents to ingest massive amounts of content in minutes, site owners need to be sure that there content is well set up for success.
This led me down an incredibly exciting path of exploration about what factors marketers will need to consider when optimizing their content for AI agents. As a result, I’ve started playing around with different use cases and functions and wanted to create some best practices that I believe we’ll all need to utilize in order to be successful.
What Is Agentic SEO?
Agentic SEO is the practice of optimizing your site for AI agents. Agentic SEO blends both content and technical SEO to ensure that AI agents can access your content, understand the context and find information that helps answers prompts. To be great at Agentic SEO will require a strong understanding of both the technical foundations of your site as well as an extremely robust understanding of how your core users will search.
It’s very likely that as adoption of AI grows and users get more and more comfortable with LLMs, they’ll start to tap into their more advanced capabilities. Models such as ChatGPT’s Agent Mode, allow users to give an AI agent commands (find a car rental in Salt Lake City, gather the best portable storage units near me). The AI agent will then act on behalf of the user searching through many different documents finding all the information they can.
At the end of all the findings and research, the AI agent will provide a comprehensive list of all the findings that they found.
This could very well be the way that information discovery happens in a 2-3 year time horizon, especially for more conversion-driven and higher cost evaluations.
AI Agents Will Need Search Indexes
The good news for SEOs that in the near-future, AI agents will likely need to rely on search indexes to gather their data and information. Indexes like Google represent the most comprehensive information discovery platform in the world and AI agents would be remiss to ignore that.
As well, they’ll likely need to use search indexes as a means of interacting with sites. Data around pricing, availability, inventory levels, shipping etc is all extremely dynamic. An agent wouldn’t be able to use it’s historical knowledge reliably enough, they’d need they information available on the larger web and owned properties. They’ll likely use search as an intermediary to do so.
The Best Practices For Agentic SEO
So now let’s start talking about what the best practices are for optimizing your site for AI agents. In general, these are going to be actually in line with more standard SEO best practices, with a few nuances that are specific to agents looking at your content.
For the context of this article, I’m going to be mostly using examples from ChatGPT’s Agent Mode as I believe that this will be the primary AI agent that users deploy in the long term. However, as of the writing of this article its very possible that Google releases their own which would have massive implications.
1. Have Strong Visibility For The Initial Agent Search Query
It’s important to understand that when an AI agent begins it’s process, it almost always uses a search as a means of doing so. That means that SEO is critical to the underlying information discovery of agentic SEO. The AI agents use search as an information discovery platform. This means that in order to be in the consideration set, you need to understand what the agent is searching for. There’s already a lot of evidence that systems like ChatGPT are using Google’s index. This means its a key link for discovery for AI agent.
In this example I asked it to find the best 16ft portable storage units in Pittsburgh, PA. Agent Mode opens up with a search of “16ft portable storage units in Pittsburgh, PA. It then evaluates the results such as UNITS, Go Mini’s PODs and more.
Notice how this is extremely similar to the order of the Google search results.
The results that already rank well in this query are immediately put in the consideration set. This means that strong rankings/visibility are one of the biggest first steps in actually performing well for agentic SEO. Without strong rankings, you won’t even be in the consideration set.
As well, throughout the process AI agents might continue to refine their searches. In this example, they changed to the search to “portable storage container Pittsburgh 16ft” and evaluated a different set of results.
At the the end of the day, this will tie back to your traditional SEO ranking tools such as Ahrefs, SEMRush, STAT etc. How you rank for the prompt will directly impact whether or not the agent will even find your content.
2. Ensure Your Site Is Accessible To The AI Agent
Once you’re in the consideration set for the AI agent, you’ll next need to ensure that it can actually read your content. This is an exceptionally common mistake that I see. A site performs well against the search but the agent can’t actually access their content.
Requiring human verification is one of the biggest hurdles to AI agent discoverability. Here’s an example of it not accessing a high ranking page because of the issue.
The AI agent even takes a screenshot of exactly what it sees when it comes across the page.
As a result, this site is blocking it’s opportunity to be included in any AI agent search. This means that they’ve lost the ability to control their own destiny. The brand company might still be included, but it will need to have a presence on third party sites that are accessible.
Some of the common reasons we see sites get blocked include:
- Human verification requirements (CloudFlare setting)
- JavaScript content
- Dynamic rendering
- No text-based content
You’ll want to test your site to see you’re able to even be included for an individual agent’s request. If not, you’ll never be able to use your site as a means of ingesting the content into AI.
3. Understand The Agent’s Evaluation Criteria
The AI agent will likely have different evaluation criteria that it infers against the source prompt. Just because the prompt doesn’t specify criteria, doesn’t mean the agent won’t look for it.
With ChatGPT Agent Mode, it will often tell you directly the types of information that it’s going to search for in a given piece of content. For example, for our “Portable Storage Unit” prompt, it looks against “price, size, climate control and regional coverage”.
Agents won’t always tell you the decision criteria, but it can be extremely valuable when they do. This will give you the context to better understand what the AI agent will consider when forming its recommendations for the end user.
4. Include Information Related To Prompts On The Page
Now that you understand the AI agent’s evaluation criteria, you need to ensure that this information is available on your page. The agent is specifically going to look for this data or other data specified by the user. What’s pretty amazing is the the agents have the ability to actually search within pages to better find the information they’re looking for.
In our example of looking for a 16ft portable storage unit, you can see that Agent Mode actually performs an on-page search for “16 ft” on a longer page that it found.
When it performs this search, it is able to find the content on this page. It extracts from just a single sentence that mentions the different sizing options available. This content is eventually included in the final output.
Having your information directly on the page is going to be the lowest friction way to directly provide your data to an AI agent.
5. Create And Link Supporting Pages
AI agents don’t only look directly at the page they first land on. They can click on existing element on the page or even perform a different search to get more information. It’s likely that they use some sort of similarity calculation to determine what the next step is.
If AI agents can’t find the information directly on your initial page, they might choose to dive deeper by looking at supporting pages. Here they didn’t the information they wanted on Two Men and Trucks initial page, so they clicked on the “Portable Storage” link.
Here’s an example where they performed a new search. AI Mode couldn’t read information on 1-800-Pack-Rat, so they decided to perform a search for their different container sizes.
They then looked the first result that came up and used that supporting page as part of the information the LLM digests.
While I wouldn’t recommend relying on it, having supporting pages with the proper information can also be accessed by AI agents. However, it can be very difficult to determine when they actually decide they need to look for a separate page vs giving up and moving on to another site. Bigger well established brands might get more leeway here.
6. Optimize Key Third Party Sites
One critical thing to mention about AI agents is that they’re not only grabbing information from you and competitor sites. They’re also able to leverage third party aggregator sites to find information about your company. Third party sites tend to be an extremely valuable source of data since they directly compare multiple options, often don’t block agents and use friendly formatting such as tables and lists.
In this example, one of the sites for our 16ft portable storage unit is blocked. However, this isn’t enough to deter the agent. Instead, it switches to Google and perform a brand search (UNITS moving and portable storage of Pittsburgh) for one of the companies:
The search is successful and the AI agent is able to pull data from a third party review that it found via search.
Here is another example from a search I performed. The original site was inaccessible to the AI agent. It’s workaround was to perform a search for an external directory. It then used the data in that third party site in the analysis.
Now ideally, AI agents are able to get the information from your site. But if you aren’t allowing the agents to access your content, you need to be hyper-aware of the third party sites that list your information and what data it’s able to pull from there.
7. Utilize Structured Content: Tables, Divs, Bulleted Lists
When analyzing the outputs of what AI agents are providing, one thing seems pretty clear. It’s much more likely to include content that’s well structured in the final response. It’s pretty common to see that heavily uses structured formats such as tables, divs, bulleted etc be the predominant source in the final output.
Probably the best example of this is in a search I did for a 7-person car rental in Salt Lake City. To start the AI agent started looking at sites you’d expect such as Kayak, Avis, Budget and more. However, none of them got pulled into the final response.
Instead, Agent Mode found a site called CARNGO. Google has absolutely demolished their organic traffic during the last year. However, they’ve formatted their content into well structured <div> tags that make it easy for ChatGPT to extract the data.
As a result, they’re actually the only result that pulled into the final output of the response. This structured content seemed to be enough for the AI Agent to completely disregard all the other information that it found.
8. Highlight Your Pricing And Discounts
At the end of the evaluation, the AI Agent will aggregate all the information it’s found. As of now ChatGPT’s AI Mode creates a table that lists all the different options that it believes are best for the user. It order the results based on what it believes the best options are.
While not always the case, this is oftentimes sorted from the cheapest option to the most expensive option. In this example for a car rental in Salt Lake City, it orders the rental options from the $71/day to $1,800/day.
I’ve even seen examples where Agent Mode takes discounts that were mentioned in site title tags (not even the page copy) and integrates them into the final pricing that it gives. If you can compete on pricing information or offer product discounts, you’ll want to ensure you’re leveraging that on your key landing pages. This can give you an edge in the final output of agentic SEO.
Conclusion
Agentic SEO represents the next frontier in search optimization, a world where we’re no longer just optimizing for human readers and traditional algorithms, but also for autonomous AI agents tasked with gathering, interpreting, and recommending content on behalf of users.
Success in this environment will hinge on mastering the fundamentals of SEO while adapting to new realities: ensuring discoverability in agent search queries, making your site fully accessible, aligning with agents’ evaluation criteria, structuring your data for frictionless extraction, and strategically managing both your owned content and key third-party listings.
The brands that embrace Agentic SEO now will gain a first-mover advantage, positioning themselves not only to be found, but to be consistently chosen by AI agents in the years ahead. As the line between search engine and autonomous assistant blurs, the winners will be those who optimize for the agents that hold the keys to tomorrow’s information discovery
