Director of Organic Search Strategy
The traditional search experience of the past several years was marked in some ways by choice and selection. User research in 2019 found that searchers distributed their attention across search results in a “pinball” pattern. Sites ranking on the first page of search results, especially the top five positions, had a reasonable chance of earning a click.
In contrast, AI search is marked by concentration and generation. AI-generated synthesis now appears at the top of many Google results, and AI chatbots can respond to information requests with text containing no links. The models powering these experiences generate fluent, confident language, a skill they maintain even when factual reliability lags behind.
It’s increasingly likely that searchers will form an understanding of your space and brand based on AI-generated information. Many things contribute to what AI generates. Of these, your website is the part most directly in your control. That is what we will focus on in this article. Third-party mentions matter too, but that’s a rich subject worth covering separately.
This article walks through how to prioritize website improvements with AI search in mind. It starts with broader fundamentals, then narrows down to suggests for where to concentrate effort. While AI search is the motivating factor behind the approach outlined here, its benefits are not exclusive to AI. Good modern SEO complements good AI SEO.
However, AI search invites a different standard of rigor in some areas, specifically around technical fundamentals, content structure and depth, and whether key ideas can stand on their own when extracted. This article highlights those distinctions.
Technical accessibility is vital. Traditional technical SEO fundamentals apply to AI SEO, but a few areas deserve extra attention. Prioritize fixing any major issues related to content in HTML, internal linking, and speed.
Key content should be in the initial HTML.
AI crawlers cannot reliably render JavaScript the way that Googlebot generally can. Screaming Frog has JavaScript reports that help surface which pages are affected. Follow the well-established best practice of serving important content in the initial HTML using server-side rendering or pre-rendering.
Pages should load quickly.
AI crawlers may have tighter timeout thresholds than traditional crawlers like Googlebot.
Make important content easy to find through internal links.
This has always been a core SEO best practice, and Google still calls it out in guidance for AI features.
Structured data might help in limited ways.
Structured data mainly supports the traditional search layer, which may be used during AI search depending on the prompt. Part 1 of this series walks through the difference between training data, web search, and how traditional search may fit in.
Core contact details like your address and phone numbers belong in highly accessible places, like on the homepage, in the footer, and on pages linked within site navigation. This is existing best practice, but it’s extra important now because phone numbers in particular are at risk for scams and friction in AI search. You can read more about this in our article about why LLMs get things wrong.
AI search has introduced new dynamics into how content gets found and used (covered in Part 1), and this section covers some new considerations. But for SEO and content teams, the core work hasn’t fundamentally changed. It’s still valuable to think about content at an intent level, build pages to address that intent with well-structured supporting ideas, and align what you publish with audience needs.
As for changes, there are two meaningful shifts worth addressing up top:
First, queries are getting longer and more conversational. Search demand for an intent is often dispersed across long-tail variants more than it used to be. For this reason, long-tail queries, including very low-volume variants, are increasingly a valuable part of understanding search demand and should be part of AI-related analysis.
Second, AI can summarize or recombine parts of your content away from original content. There is a heightened need to to consider how ideas are presented on the page and whether they hold up when extracted on their own.
The rest of this section covers how to approach content through this lens, starting with content standards for AI, then aligning on priorities, finding AI-relevant topic ideas, and types of queries to focus on.
There is a lot of good information online about creating AI-friendly content and the types of content that models tend to cite, such as from iPullRank, Semrush, and Growth Memo.
The idea of “chunking” is one of the more-discussed areas in AI-focused content right now. As discussed in Part 1 of this series, many models pull specific sentences or passages (chunks) when responding to some prompts. Chunk boundaries vary. This article offers a helpful example of this practice at work.
Officially, Google has advised against breaking content down into bite-size chunks for ranking purposes. I agree that it’s not a good idea to create content solely for machine consumption.
At the same time, machines must be able to access, interpret and extract value from what you publish in order for your work to get found organically. So for me, the aim is to write for humans first, with an acute awareness of how the machines work.
With that in mind, below are characteristics of AI-friendly content that come up consistently across well-respected sources.
The standards below are a strong starting point.
Establish well-organized structure reflected through clear, concise headings.
LLMs often use headings to understand what a section is about. If your heading is vague or confusing, the content underneath may be misclassified or skipped.
Write self-contained paragraphs.
A self-contained paragraph communicates one idea that makes sense when quoted alone. Avoid paragraphs that rely on the heading for meaning, mix multiple ideas, or use orphaned pronouns (like “this”) with unclear references.
Be specific and back up claims.
“Show don’t tell” is a classic principle of strong writing, and in AI search, publishing content that is clear, specific, and backed with data can help move the needle in your favor.
Keep terminology consistent across the site.
Small wording differences can lead AI systems to treat related statements as separate. For example, if one page says “per seat” and another says “per user license,” a model may not reliably infer that they mean the same thing.
With that context in mind, here’s how to approach content on your website through the lens of AI.
AI-friendly content is clear and specific. At the same time, official data on user prompts and AI outputs is limited, particularly for AI chatbots.
Put differently, we’re contending with an environment where precise content seems valuable for people and emerging search technologies, but also we have limited first-party data to guide content development attuned to evolving user needs.
While first-party data on AI search behavior is still limited, clarity about what you offer and what your audience needs is something you can build from many existing sources. When directing focused effort towards AI search, it helps to be aligned on fundamentals like:
What are you focused on? A specific product, line, or the full brand?
Who are you trying to reach? Which audience or segment matters most?
Once aligned on priorities, the next step is building an inventory of topics your site should cover. Pull from multiple sources, which might include:
Sales/support questions, and customer reviews
These reflect friction points, decision factors, and feedback, all in your audience’s own words
Google Search Console & Bing Webmaster Tools
Citations in AI Overview and AI Mode count as an impression, and Bing now has AI data
Manual prompt testing
Testing what AI tools say can be directionally useful in some cases, and can help reveal incorrect facts
AI visibility tools
These tools can be directionally helpful, but have limits discussed at the end of this article
Traditional keyword research tools
These can signal likely search demand for a topic that could carry over to AI prompts
A note on search volume as a signal
Traditional search volume isn’t a direct measure of AI prompt activity, but it can be a proxy for search interest surrounding a topic. For some tools, “0” search volume can actually represent up to 9 estimated searches per month. Keep this in mind if you see dozens of related “0” variants for queries highly relevant to a target audience. Actual search demand could be higher than estimates suggest.
AI search has changed where effort can pay off. It’s worth keeping potential payoff in mind when deciding which topic spaces likely warrant closer attention and dedicated time.
Informational, top-of-the-funnel content has been affected dramatically by AI search, particularly due to AI Overviews. These queries get fewer clicks than they used to. AI visibility can also be challenging to secure, especially for well-established facts.
In light of this, investment in top-of-the-funnel content tends to make the most sense when a topic is closely tied to brand identity, for areas when freshness matters and you have something new to share, and for long-tail concepts highly specific to your audience.
Put effort towards topics that support consideration and decision-making. First, the people behind these queries are comparing options and assessing fit. They could convert. Second, if AI is generating answers about your brand or key non-branded topics, it’s risky to be absent from that conversation. Third, long-tail queries in this space can offer quick wins in both AI and traditional search by addressing these specific topics on your site.
Some topics under this umbrella worth exploring include:
Feature and functionality, including phrases like: “does it have,” “can it,” “does it support”
Comparisons (“vs”), who you are best for and why (“is [company/product] good for”)
Cost and value: tradeoffs, comparative value (“worth it”).
How you prioritize will vary per team. Some prefer informal collaboration and others work through a structured scoring system. Resource constraints may also impact how quickly you move. But for content-related work, these two factors should be part of your decision-making:
Topic Importance. How central is the topic to your brand and business goals? It is a key part of how your audience evaluates and makes decisions? If your content is weak or missing, what’s the likely cost? Data on AI visibility, referral traffic, and other metrics sharpen this assessment, although AI measurement comes with its own challenges. The next section covers this.
Content Readiness. Does your existing content on this topic reflect AI-friendly content standards for clarity, structure, and specificity? Is there a clear destination page for the intent on your site, or is spread thin across multiple pages?
Below is a prioritization matrix that shows one way to act on the answers to those questions.
Create now. There is a high-impact topic not addressed on your site. Create a new page or add a section to an existing one.
Fix now. High-impact content that exists but doesn’t meet content standards. Restructure, clarify, and improve specificity where needed.
Monitor. High-impact pages that already meet content standards. Monitor.
Backlog. Low impact pages not up to standard. Fix later.
No action required. Low-impact pages up to standard.
Tracking AI search is an evolving challenge. Different models produce different responses, and outputs vary by session. Due to the variability of LLM outputs, AI visibility tracking should not be interpreted like traditional rankings. You cannot track “position” in AI chatbot responses the way you can track keyword rankings today. Instead, we must monitor and analyze signals from multiple sources.
For AI visibility, Google Search Console reports AI Overview and AI Mode citations as impressions, and Bing Webmaster Tools began reporting AI performance data in early 2025. AI monitoring tools can potentially add directional context in some circumstances, but their reliability is still an open question. SparkToro’s recent research on inconsistency in AI visibility data is worth reading on this. Manual prompt testing can help selectively, like catching obvious inaccuracies or distorted positioning.
For referral traffic, web analytics tools like GA4 report visitors from AI platforms, though many AI users do not click through, so this will likely represent a small share of overall traffic. You can also use web analytics to help connect AI-originating activity to conversions or revenue.
For on-the-ground insight, it’s worthwhile to ask how customers found you in demo booking forms or sales conversations. Feedback like this can surface patterns that tools alone might miss or only partially capture.
AI search is still evolving. Precise tactics will shift. This article focused on the fundamentals worth investing in, which support AI visibility, complement traditional SEO, and help improve overall user experience.
