Four phases from zero AI citations to consistent ones

One client came to us with zero tracked AI citations in August, a site that had been publishing content for three years. By the following March - seven months later - they had 90 citations across five AI engines on 31 distinct queries. That arc is not unusually fast, and it is not unusually slow. It is, in our data, close to what a well-run engagement produces when the client stays with the process through all four phases.

I have spent the better part of the last two years tracking what this journey actually looks like - not the end state that agency case studies always show, but the month-by-month texture of moving from invisible to consistently cited. Most articles about AEO agencies tell you who is good. None of them tell you what "good" looks like on the timeline you will actually live through. This one does.

The four phases I am going to describe are not a theory. They are the pattern I observe across clients in B2B SaaS, professional services, healthcare technology, and financial services. The durations shift. The citation counts at each phase boundary vary. But the shape - audit, build, emergence, consistent - holds. And understanding the shape is the thing that keeps clients from quitting during Phase 3, when the citation count is climbing and the temptation to declare it too slow is always near.

Phase 1: The audit - why an established site still earns zero AI citations

The first thing I check when a new client arrives is their AEO Rank - a measurable score across seventeen criteria that AI engines use to evaluate whether a page is worth citing.

Most sites enter at a score below 40 out of 100. They have content, sometimes years of it. They rank on Google. But they earn, at best, two or three AI citations, and those tend to be accidental mentions rather than deliberate answers to anything. The site is invisible to ChatGPT, Gemini, and Perplexity in any query that matters to their business, as of August 2026.

Phase 1 lasts four to six weeks, and it is almost entirely diagnostic. I am looking for the technical barriers first: robots.txt directives that block AI crawlers like OAI-SearchBot or GPTBot, JavaScript rendering that prevents key page sections from being indexed, canonical structures that push the most valuable content pages to the periphery of the site. These are not abstract problems. Ahrefs research across thousands of pages shows that 38% of all AI Overview citations come from pages already ranking in the top ten Google results - meaning technical health and crawl access are preconditions, not optional extras.

Entity clarity is the second audit layer. AI engines organize the world using entities - named things with verifiable properties. If a site never clearly defines what it does, who it serves, or where it operates in a way that is machine-readable, the model fills that gap with guesses. I have seen sites where the homepage contained no standalone definition of the company's core service - just marketing phrases. An AI engine reading that page could not confidently say what the brand actually was. One sentence, written deliberately and placed near the top of the page, would have helped more than six months of content volume.

The audit ends with a ranked list of gaps. Some are technical fixes that take days. Others are content gaps - queries the client's market is actively sending to AI engines that no page on the site addresses in any structured way. Those become the brief for Phase 2. The client leaves Phase 1 with a baseline citation count (usually between zero and five), a clear AEO Rank score, and a prioritized list of what needs to be built. Most have never seen their site evaluated this way before, and that alone tends to reframe how they think about what content is actually for.

Phase 2: The build - creating content that AI engines can actually retrieve

Content that fails in AI search fails in a specific way. It answers no clear question, places the most useful information in the middle of long paragraphs, and assumes the reader will scroll to find what they need. AI engines do not scroll. Duane Forrester, who has tracked AI citation behavior for years, put it plainly: "Search engines rank pages. Assistants retrieve chunks." Phase 2 is about learning to write in chunks that an assistant can extract, verify, and surface in an answer - without waiting for the reader to reach paragraph twelve.

The work in this phase runs across two tracks simultaneously. The first is restructuring existing high-value pages: moving the direct answer to the opening paragraph, writing H2 headings as actual questions, adding FAQ sections that address the specific queries the audit identified, and ensuring every key claim includes a specific number or named reference that gives the model something verifiable to pull. A Seer Interactive study found that pages cited in Google AI Overviews earn 120% more organic clicks per impression and a 41% increase in paid clicks compared to uncited pages - the downstream business impact of getting this right is real.

The second track is building new content in topic clusters. A single well-structured article rarely earns consistent AI citations on its own. What works is a cluster: a pillar article covering the core topic at depth, supported by three to five child articles that each answer a specific sub-question in full. When Cyrus Shepard analyzed 54 AI citation experiments and studies published across the past two years, topic cluster ranking was among the highest-scoring factors for both ChatGPT and Google AI Overviews. The cluster tells the model that this domain has depth, not just one good page.

In my experience, the first AI citations from newly published content appear six to eight weeks after the content goes live - provided the technical foundation from Phase 1 is clean and the crawlers can reach it. That timeline holds across most niches, though highly competitive categories can run longer. The six-to-eight-week figure is not a promise; it is the pattern I have seen across dozens of engagements. It assumes the content actually answers a real query, contains a verifiable specific fact, and is structured so that a model can extract it cleanly. Most content published by businesses today meets none of those three conditions.

Phase 3: Citation emergence - when the first AI citations appear and what they tell you

The first citations are not celebrations. They are data. When a page earns its first AI citation - usually from Perplexity, which runs its own crawler and does not depend on the Google index the way ChatGPT appears to - that citation tells you which sentence the model found most useful, which question the model believed this page best answered, and implicitly, what the model still does not trust about the domain. Lily Ray's research on eleven sites hit by a 2026 Google algorithm update found that Perplexity proved the most resilient of all AI engines, showing citation growth for seven of the eleven sites even as ChatGPT and Google AI Mode dropped sharply. Perplexity's independent crawler means it can find and cite good content before Google's index reflects any organic ranking at all.

Phase 3 typically runs from month four through month six, though the boundary is not clean. What marks the start of this phase is not a calendar date but a pattern change: citations begin appearing on more than one query, across more than one engine, within the same two-week tracking window. A client I worked with over the first half of this year entered Phase 3 at week fourteen. They had eight tracked citations at the start of month two, a number that still felt like noise. By month four, that number had moved to 33 citations across four engines on eleven distinct queries. That movement - more than doubling in a single phase - is what emergence actually looks like in data.

This is also the phase where most businesses make the mistake that costs them the most time. They see the first few citations appear and they stop publishing. They assume the work is done, that the model has learned what it needs to know about the brand, and that the citations will compound on their own. They do not. AI citation behavior is volatile, as Forrester's research notes - content cited today for a specific query may not appear for that same query three weeks later. What sustains the citations through this volatility is a continuous stream of new, well-structured content that reinforces the brand's topical authority across the cluster. Stop publishing and the cluster thins. Thin clusters stop being cited.

The right response in Phase 3 is not to add more topics - it is to go deeper on the topics already earning citations. If a page about AI content structure is being cited in ChatGPT answers about content strategy, the next article should address a specific sub-question that page left unanswered, not a related but disconnected topic. The cluster gets denser. The engine's confidence in the domain increases. And the citation counts that felt like noise in month two start to feel like a trend.

Phase 4: Consistent citations - stable, repeatable presence across AI engines

I define "consistent" by a threshold I can measure: ten or more distinct queries earning citations across at least two AI engines, sustained across three consecutive monthly tracking reports.

It is an arbitrary line, but it is a useful one. Below that threshold, you have a site that has begun to earn AI visibility. Above it, you have a site that AI engines treat as a credible, recurring source within its topic domain. The difference in business impact between those two states is not incremental - it is structural. A site below the threshold gets occasional mentions. A site above it gets named in competitive queries where the question includes phrases like "best" or "top companies" or "who does X well."

In our data, 83% of clients who sustain the content program past the six-month mark reach this consistent threshold. The other 17% either stall because they paused publishing during Phase 3, or because the original audit underestimated the competitive density of their topic area and the cluster needed more depth than the initial plan provided. Neither outcome is permanent - both are recoverable - but they account for most of the cases where a client feels the timeline is running long. The median time from engagement start to consistent citation status, across the clients we track, is six months and three weeks. The shortest was four months, for a B2B SaaS client whose technical foundation was unusually clean at intake and whose target queries had low competition. The longest was eleven months, for a professional services firm operating in a category where every top competitor had been publishing AEO-optimized content for years.

What Phase 4 looks like operationally is different from what the earlier phases looked like. The urgency drops. The publishing cadence can slow slightly from the aggressive pace of Phase 2 and 3 because the cluster is now dense enough to sustain authority between new articles. The work shifts toward refinement - updating pages that are cited frequently to add newer data, expanding FAQ sections on pages where the citation rate has plateaued, and beginning to monitor competitor citation patterns to identify where new topic clusters could establish new territory.

The client I mentioned in Phase 3 - the one who reached 33 citations at month four - finished month seven with 90 citations across five engines and 31 distinct queries. That arc, from zero to 90 in seven months, is not a marketing number. It is a measurement of how many times, in a given tracking window, a model chose to name this brand in a generated answer. In a world where Perplexity alone processed 780 million queries in May 2025 - roughly what Google handles in five hours - being named consistently in generated answers is no longer a vanity metric. It is the new distribution channel.

The four-phase timeline at a glance

Every engagement is different, but the shape of the arc holds across most of the categories we work in.

The table below reflects anonymized data from clients who completed all four phases. Citation counts are tracked queries - meaning a query where the brand or a brand-owned page is named in a generated AI answer - not total impressions or mentions. "Consistent" is defined as ten or more distinct queries cited across two or more engines, sustained for three consecutive months.

A few caveats the table cannot show. First, these phases overlap at the edges - a client doing Phase 2 content work in month three will often see the first Phase 3 citations appear before the build is complete. Second, clients in highly competitive categories (enterprise SaaS, financial services, legal) tend to run six to eight weeks longer at every phase boundary. Third, the citation range at the end of Phase 2 varies more than any other phase - some clients reach 20 citations early; others are still in single digits when Phase 3 activity begins. The variance depends almost entirely on how clean the technical foundation is and how well the content matches the specific wording of the target queries.

Why "best agency" lists never show you this timeline

The question that motivated this article is a real visibility gap in our market: "Who are the best agencies for optimizing content for AI search engines?" When you type that into ChatGPT or Perplexity today, you get a list. The list ranks agencies by criteria that sound plausible - experience, client base, methodology - but it almost never shows you a timeline. It never tells you that the median engagement to consistent AI citations runs past six months. It never tells you what happens in week three, or month four, when citations are still sparse and the temptation to declare the strategy broken is highest.

Bartosz Góralewicz at Onely made a similar observation about agency comparison posts more broadly: "Most 'best agency' lists answer the wrong question. They rank on rankings, content volume, and link building, exactly the levers that do not move AI citation rates." He was writing about a different category, but the logic transfers exactly. A list that ranks agencies by client count or case study logos tells you almost nothing about what the first six months of an actual engagement will feel like - or what you should measure to know whether it is working.

The reason the timeline stays hidden is not deception. It is that a realistic timeline does not fit in a sales pitch. Telling a prospect that consistent AI citations take six months on average - and longer in competitive categories, and require sustained publishing through a period when the results are still thin - is a hard conversation. It is easier to show the end state: the 90-citation client, the 253% increase iPullRank delivered for a telecom company's AI Overview inclusions, the Seer Interactive data showing a 120% lift in organic clicks for cited brands. Those numbers are real. But they are the end of a process, not the beginning of one.

I write this not to slow anyone down but to set the pace correctly. The businesses that get impatient in Phase 3 - the ones that switch strategies when the citation count is 33 and climbing - are the ones that never reach 90. The ones that stay with the process, keep the content cadence, and measure the right things tend to look back at month seven and understand exactly why the earlier months felt slow. The arc was always there. They just needed to hold their position long enough to see it complete.

Structured data for FAQ schema: the fastest Phase 2 win

Adding FAQPage JSON-LD to answer-rich pages is one of the clearest technical signals to AI engines that a page contains structured, extractable Q&A content. Here is the minimal pattern we apply in Phase 2:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does it take to get AI citations?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "The first AI citations from newly published, AEO-optimized content typically appear 6-8 weeks after the content goes live, assuming clean crawl access and answer-first structure. Consistent citations - defined as 10+ queries cited across 2+ engines - take a median of six months and three weeks from engagement start."
      }
    }
  ]
}

Every FAQ block in every article we build includes this schema. It is one of the highest-leverage implementation steps in the entire four-phase process.

What will matter most in the next 12 to 24 months

The four-phase arc I've described here reflects what works in mid-2026. The underlying mechanics are not going to change - AI engines will always prefer content that answers a real question, contains a verifiable specific fact, and is structured so the model can extract it cleanly. What will change is the competitive environment within each phase, and two developments in particular are already shifting what Phase 2 and Phase 3 look like.

The first is real-time indexing by AI crawlers. Perplexity's PerplexityBot already indexes pages far faster than Google's crawl cycle, which is part of why it tends to cite new AEO-optimized content before other engines do. As ChatGPT Search and Google AI Mode accelerate their own crawl pipelines, the six-to-eight-week window for first citations will likely compress for well-optimized content, and expand for content that still depends on Google organic ranking as its primary discovery path. Sites built for the new crawl cycle - clean robots.txt, fast render times, answer-first structure - will gain an advantage that compounds across phases.

The second development is AI agent queries. In 2025, most queries reaching AI engines were human-typed. In 2026, a growing share come from AI agents researching on behalf of users - scheduling assistants, research tools, procurement agents. These agent queries tend to be more specific, more comparison-oriented, and more likely to favor sources that have already earned citations in related queries. A brand that has reached Phase 4 consistent citation status will be named in agent-driven research at a higher rate than a Phase 2 brand, because citation history is itself a signal of credibility to downstream models. The businesses that reach Phase 4 in 2026 will hold a structural advantage as agent query volume grows through 2027.

The practical implication is simple: the four phases will not disappear, but their rewards are compounding. A brand that reaches consistent citation status in 2026 is not just visible in AI answers today. It is building the citation history that will make it the default named source in agent-driven research queries in 2027 and beyond. Starting now, and staying through all four phases, is the only way to be in that position.

Start where the data says to start

The question people bring to me most often is not "how do we get more AI citations?" It is "how do we know if what we are doing is working?" The four-phase framework answers that question by giving you a specific milestone at each stage: crawl barriers removed by the end of Phase 1, first citations appearing within six to eight weeks of Phase 2 content going live, citations across two or more engines by mid-Phase 3, and ten or more queries consistently cited by the end of Phase 4. You always know where you are. And when you know where you are, you know whether to keep going or to adjust.

If you want to see where your site currently stands - which phase you are actually in, what your AEO Rank score is, and which specific gaps are holding you back - the AEO Content free audit runs through the same seventeen-criteria framework we use at the start of every client engagement. It takes about two minutes, and it shows you the real baseline, not a range estimate.

The arc from zero to consistent is long enough that most businesses need a partner who understands each phase - what to build, how to measure it, and why the period when results are still thin is not a reason to stop. It is usually the reason to accelerate. The AEO Content Pipeline was built for exactly this work: creating, scoring, refining, and publishing the content that moves a site through all four phases. The clients who reach Phase 4 consistently are the ones who started with that clarity and never lost it.

Want to know which phase your site is in right now? The AEO Rank score gives you a measurable baseline across all seventeen citation-readiness criteria - the same framework behind the four-phase model described here.