How to build a content program for AI search from scratch

Most of what gets published about AI search optimization assumes you already have a content library. The advice is familiar: audit your existing blog, restructure your highest-traffic posts, layer citation optimization onto your SEO program.

For a company with 200 published articles, that’s useful. For a company with zero, it’s useless.

The traditional fallback is “build for SEO first, add AI search later.” That sounds reasonable until you look at the data. The overlap between organic rankings and AI citations is far smaller than the SEO-first worldview assumes, and the cost of retrofitting a content library for AI search after the fact is real.

A company starting from scratch has an option that existing programs don’t: design for AI citation from day one, without inheriting the structural debt that makes retrofitting painful.

Why every AI search guide assumes you have content

Open any article about AI search optimization and count the verbs. Audit. Restructure. Retrofit. Optimize. Repurpose. Every one of those assumes something already exists. The default starting point is a content library with traffic, rankings, and an established domain.

The assumption isn’t accidental. The AI search optimization space grew out of the SEO industry, and the SEO industry built its frameworks around existing websites.

Practitioners wrote the first wave of AI search advice to solve their own problem: how to make content that already ranked in Google also get cited by ChatGPT, Perplexity, and Google AI Mode.

For a company with no content, no organic traffic, and no domain authority, the advice to “audit and restructure” is a dead end. And the fallback (“build traditional SEO first, add AI later”) may actually be the slower path.

Moz’s analysis of 40,000 queries found that 88% of AI Mode citations go to URLs that don’t appear in the organic top 10. Only 12% showed a strict URL match with organic results, and only 20% matched at the domain level. The playing field for AI citations looks nothing like the one for organic rankings.

The data that changes the starting-from-zero equation

Three pieces of research reframe what it means to start a content program without existing assets. Each one challenges a core assumption of the “build for SEO first” approach.

Data point	What it means for existing programs	What it means for new programs
88% of AI Mode citations go to URLs outside the organic top 10 (Moz)	Organic rankings don’t guarantee AI citations. Existing content needs structural changes to earn them.	Starting without organic rankings is less of a handicap than expected. The citation landscape is more open than the SERP.
ChatGPT citations grow from 10% on day 1 to 42% by day 30 on a DA 84 domain (Semrush)	New pages on established domains can earn stable citations within a month.	Citation timelines are real but not prohibitive. New domains face longer timelines (more on that below), but the pattern is acceleration, not stagnation.
65-85% of ChatGPT prompts don’t match traditional search keywords (Semrush)	Traditional keyword research misses most of the queries AI users actually ask.	A new program doesn’t need to start with keyword research. It can start with the questions AI users are asking, which look nothing like traditional search queries.

The Moz finding is the one that matters most. If AI search engines selected sources the same way Google organic does, starting from zero would be a serious disadvantage. But the 88% non-overlap means AI models are evaluating content on criteria that have little to do with organic rankings: topical depth, answer completeness, entity clarity, and source credibility. A new content program can compete on all of those without a single page ranking on Google first.

How to design a content program for AI search from day one

Five design decisions separate a content program built for AI citation from one built for traditional SEO and retrofitted later. Each decision is specific enough to act on this week.

1. Start with one topic cluster, not ten keywords

Traditional SEO advice says target your highest-volume keywords first. Spread your bets. Cover as many search queries as possible with your first batch of content.

For AI search, that advice gets the sequencing wrong.

AI models don’t cite brands that rank for individual keywords. They cite brands that demonstrate expertise across an entire topic.

Keyword Insights found that pages with high topical authority gain traffic 57% faster than pages without it, and the same principle applies (with even more force) to AI citations.

Pick one topic cluster in your category and build 10 to 15 pieces that cover it from every angle: the definition, the common mistakes, the how-to execution, the comparison with alternatives, the measurement approach, the case for and against.

A B2B SaaS company selling project management software, for example, would be better served by 12 articles covering “resource capacity planning” from every angle than by 12 articles each targeting a different high-volume keyword.

The reasoning is straightforward. Traditional keyword volume doesn’t predict which queries AI users ask, and AI models evaluate topical coverage at the cluster level, not the page level. Depth in one area beats breadth across many.

2. Sequence content for citation momentum, not traffic volume

The order in which you publish matters more for AI search than for traditional SEO.

With SEO, you can publish in any order and let Google crawl and rank each page independently.

AI models work differently: they build confidence in a source over time, and the sequence of what you publish affects how quickly that confidence develops.

Start with definitional content. What is X? What does X mean for your category?

Then build out problem-diagnosis content: why X fails, what goes wrong with X, common mistakes in X.

After that, publish how-to execution content, and finish with comparison and evaluation pieces.

The sequence mirrors how AI models assess source credibility. A brand that explains the fundamentals clearly earns the right to make advanced claims. A brand that jumps straight to “10 best tools for X” without ever explaining what X is looks like an affiliate site, not an authority.

Semrush’s citation speed data supports the pattern. On a DA 84 domain, Google AI Mode cited 36% of new pages on day 1, but ChatGPT started at just 10% and grew to 42% by day 30. ChatGPT builds trust slowly and rewards consistency.

The right content sequence accelerates that trust-building, and understanding why AI search retrieves content but doesn’t always cite it helps you diagnose gaps in the sequence.

3. Design every page as a standalone answer

AI search engines don’t send readers to your site to browse. They extract specific answers and cite the source.

Every page in the program needs to contain at least one self-contained answer that makes sense without reading anything else on your site.

The structural requirements are specific. Each page needs a clear, direct answer to the question it targets, placed high on the page.

The answer should be self-contained: a reader (or an AI model) should be able to extract it without needing context from the introduction, the heading, or another article. Supporting evidence, examples, and nuance come after the direct answer, not before it.

For a program built from scratch, these structural requirements for AI search citations become part of the content template from page one. Every brief, every outline, every draft review checks for standalone answers.

Existing programs have to retrofit hundreds of posts that were written for a different reading model. A new program avoids that debt entirely.

4. Build measurement before you build content

Companies publish 20 or 30 posts, then realize they have no way to measure whether AI search engines are citing them.

CommonMind’s research found that 59% of B2B SaaS companies can’t see AI-referred traffic in their analytics. Most of them built measurement as an afterthought.

A new program has the advantage of setting up tracking before the first post goes live. That means configuring analytics to identify AI referral sources (ChatGPT, Perplexity, Google AI Mode), setting up citation monitoring to track where and when your content gets cited, and establishing a baseline before any content exists so you can measure change from zero.

The principle is simple: if you can’t measure AI citations and AI-referred traffic from day one, you’re flying blind for the first six months. By the time you add measurement, you’ve lost the baseline data that would tell you what’s working.

5. Plan for dual-channel from the start

A content program built for AI search doesn’t ignore traditional SEO. The structural requirements for AI citation (standalone answers, topical depth, entity clarity, internal linking) overlap with what Google rewards. The difference is sequencing and emphasis.

A traditional SEO-first program optimizes for Google rankings and adds AI search as an afterthought. A dual-channel program designs content architecture for AI citation first and gets SEO performance as a byproduct.

The distinction shows up most clearly in topic selection, where clusters replace keyword lists. Content structure shifts too: every page is built for extractable answers rather than long-form narrative. And measurement expands from traffic analytics to include citation tracking.

The practical result: every page in the program should work for both channels, but the design decisions should favor AI citation when the two channels conflict.

When Google wants a 3,000-word guide and AI models want a clear, extractable answer in the first 200 words, the answer-first structure wins. Google doesn’t penalize it, and AI models reward it.

What does dual-channel look like in practice? The editorial calendar targets topic clusters instead of keyword lists. Content briefs require a standalone answer in the first 200 words of every post. Internal linking connects cluster pages to reinforce topical authority for both Google’s crawlers and AI models’ entity graphs.

On the measurement side, the dashboard tracks organic traffic and AI citations side by side, so you can see which channel is responding to which content.

The cold-start problem and how to manage it

An honest caveat: the Semrush citation speed data comes from an established domain with a domain authority score of 84.

A brand-new domain with no backlinks, no brand mentions, and no publishing history will face longer timelines. As the Semrush team noted, “If your domain authority is high, you may have a better shot at getting cited quickly.”

A slower timeline doesn’t make the from-scratch approach wrong. It makes the timeline longer. Here’s how to manage the cold start:

1. Publish on third-party platforms in parallel. Guest posts, LinkedIn articles, and industry publications build brand mentions that AI models can find even before your own domain has authority. Timeline: weeks 1 through 8 of the program.
2. Earn backlinks through the content itself. Original data, specific frameworks, and contrarian analysis attract links naturally. The topic cluster approach produces link-worthy content as a byproduct of covering a topic deeply. Timeline: months 2 through 6.
3. Build topical authority before expecting citations. Ron Morgan at Horizon Marketing frames the timeline well: domain authority takes years, but topical authority takes 6 to 18 months of focused cluster development. A focused smaller brand can outperform a large company on specific topics through depth alone. Timeline: months 6 through 18.
4. Accept the slower start for a stronger foundation. Norbert Hires made a fair point when he argued that the right question about AI search optimization isn’t “should I do it?” but “does it make sense for my specific company and market?” For companies in categories with meaningful AI query volume, the from-scratch approach is a 6-to-18-month investment, not a quick win.

When starting from scratch beats retrofitting

DerivateX benchmarked 50 B2B SaaS companies on AI search visibility and found that 44% scored poorly.

CommonMind’s survey found that 93% of B2B SaaS companies say AI visibility is critically important, but only 14% have a mature strategy.

Both findings describe the retrofitting problem. Existing programs are dragging legacy content architecture built for a different era of search, keyword strategies based on volume data that misses 65-85% of AI queries, and measurement systems that were bolted on after the fact.

Fixing all of that while continuing to publish new content is expensive and slow.

A new program avoids every one of those costs. A new program starts without legacy architecture to untangle, keyword strategies to rewrite, or measurement gaps to backfill. The disadvantage is real (no existing traffic, no domain authority), but the advantage is equally real: no technical debt.

The DerivateX data also shows what the payoff looks like. AI-referred traffic converts at 14.2%, compared to 2.8% for Google organic. A content program that earns AI citations early, even at modest volume, produces revenue impact that justifies the investment in architecture and measurement.

The from-scratch playbook I’ve outlined here gives you the design framework. Start with the architecture, get the measurement right, build depth before breadth, and let the citations compound.

For companies deciding between hiring an SEO agency to build a traditional program or investing in an AI-first content architecture, the math increasingly favors the latter.

The 14.2% conversion rate on AI-referred traffic means fewer visits need to produce the same pipeline impact. And the structural work you do for AI citation (topical depth, standalone answers, entity clarity) makes your content better for every channel, including the organic search results you would have targeted anyway

Commonly asked questions about building a content program for AI search