Written by: Arjun Karnik, Growth Marketing Specialist
Key Takeaways
- Generative engine optimization for ChatGPT focuses on structuring content so AI systems can locate, parse, and cite pages inside machine-generated answers rather than chasing traditional rankings.
- 71% of B2B buyers now use AI chatbots for research, and citations in AI answers have become a direct vendor-selection event as zero-click rates reach 68%.
- The 7-step playbook covers running defensive audits first, then fixing technical crawlability, mapping fan-out queries, aligning content to buyer language, publishing at machine cadence, installing decay tripwires, and tracking citations.
- Pages can lose 78–99% of impressions in two months without updates, so automated refresh systems are essential for maintaining citation positions.
- See how the test lab maps this shift to your content library and identify where your current content already qualifies for citations.
The Market Shift: Buyers Ask, Machines Answer
The distribution shift is already complete for most B2B buyers. G2 surveyed 1,076 B2B software buyers across North America, EMEA, and APAC in March 2026 and found that 71% use AI chatbots for software research. 69% chose a different vendor than the one they had planned on, and 33% bought from a vendor they had not previously heard of. Being cited in an AI answer is now a vendor-selection event, not a vanity metric.
The click is disappearing where answers appear. Pew Research Center tracked 900 US adults across 68,879 Google searches in March 2025 and found that when an AI summary appeared, users clicked a traditional search result in only 8% of visits, against 15% when no summary appeared. Similarweb clickstream data shows the zero-click rate for Google searches reached 68.01% in January through April 2026, up from 60.45% in 2024.
The audience scale makes this shift unavoidable. OpenAI reported 900 million weekly active ChatGPT users in February 2026. At Google I/O in May 2026, Sundar Pichai put AI Overviews at over 2.5 billion monthly active users and AI Mode at over 1 billion monthly active users within its first year. AI assistants have become the primary research environment most buyers already use.
Inside a business that has done SEO properly for years, three signals confirm the shift. First, impressions climb while clicks fall, the Search Console scissors, because buyers get answers without leaving the AI interface. Second, traffic arriving from chatgpt.com converts like a referral, not like cold search traffic, since the AI answer has already pre-qualified the visitor. Third, sales calls start further down the funnel because the buyer was pre-educated by an AI answer before anyone from the company joined the conversation.
Map this shift to your specific content library in a live walkthrough and see where AI already pulls from your pages.
Generative Engine Optimization in Practice
The following seven steps describe the repeatable system Arjun Karnik runs on his own site, documented in public with Search Console data, decay curves, and controlled fan-out tests. Each step sets up the conditions for the one that follows.
Step 1: Fix Technical Crawlability for OAI-SearchBot
Nothing downstream works if the retrieval layer cannot read the site. ChatGPT Search relies on Bing indexing plus its own OAI-SearchBot crawler for real-time retrieval, so unblocking OAI-SearchBot and confirming Bing index status become the primary technical requirements for citation eligibility. Technical accessibility requires explicit allow rules in robots.txt for GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended, and GrokBot rather than relying on wildcard defaults.
Most current AI crawlers do not execute JavaScript. Onely’s research found that most text fragments in Google AI Overview responses came from the HTML body rather than JavaScript-dependent content. Core content must be available in view-source. Beyond making content accessible, you also need to make it machine-parsable, which is where schema markup becomes critical. Schema markup is a structural requirement, not an enhancement: pages carrying FAQPage schema appear in Google AI Overviews 3.2× more often than pages without it. Priority schema types are Article, FAQPage, HowTo, and Organization.
Step 2: Map Fan-Out Queries Directly from ChatGPT
A single buyer prompt does not produce a single lookup. Google explicitly describes query fan-out in its documentation for AI Overviews and AI Mode, stating that both may issue multiple related searches across subtopics and data sources to develop a response. Optimizing for the visible prompt while ignoring the fan-out means optimizing for the wrong surface entirely.
The solution is to reverse-engineer the machine’s actual retrieval behavior. Fan-out queries are extracted directly from ChatGPT rather than inferred from keyword tools, because the target is the machine’s questions, not the human’s typed query. In a controlled test on Arjun’s own site, pages rewritten to match extracted ChatGPT fan-out queries earned citations while control pages did not. Legacy tools like Semrush and Ahrefs rely on historical data and daily caps, which miss new long-tail queries that AI surfaces answer.
Step 3: Align Every URL, Title, and H1 to Buyer Language
Buyer language alignment turns existing expertise into content the machine can match to real questions. In Arjun’s own test, a page titled “What is GEO” was relabelled “How to Get Your Business Recommended by AI Search,” with the slug, title, H1, and H2s all realigned to buyer questions. Citations followed within weeks of that specific change.
An Ahrefs study of 75,000 brands found brand web mentions correlate at 0.664 with ChatGPT citation likelihood, compared to only 0.218 for backlinks. The machine is matching questions to answers, not weighing domain tenure. Buyer language in URLs, titles, and headings provides the alignment mechanism.
Step 4: Publish at Machine Cadence with AI Growth Agent
One person cannot publish and refresh at the cadence this channel requires. On Arjun’s own site, an AI article engine deployed on a subfolder via AI Growth Agent (a platform he uses and discloses as a partner) runs at 5 to 8 autonomous actions per day, combining new articles with updates to existing ones. The GEO subfolder went from zero to the only source of new impressions on the entire domain in 60 days, measured in Google Search Console. New articles reached thousands of monthly Google impressions within weeks.
Adding statistics increases AI citation visibility by around 31–33% and adding quotations by around 41–43%, according to the Princeton GEO study. Structure also matters: query language in URLs, titles, and H1s, schema on everything, and answer-first formatting that the retrieval layer can parse.
Step 5: Install Impression-Decay Tripwires for Automatic Refresh
Pages decay quickly without a refresh loop. In Arjun’s own tests, pages can drop 78% to 99% in two months without updates. That decay stays invisible unless the site is instrumented for it, and by the time it shows up in a monthly report the citation position is already gone. Seer Interactive analyzed 47,097 AI citations across 7,683 pages in ChatGPT, Gemini, and Perplexity between March and June 2026, and found that 75% of cited pages had been updated within the last year, with pages cited consistently across all four months averaging under six months since their last update.
Impression-decay tripwires in AI Growth Agent monitor Search Console performance and automatically queue an update when a page starts falling. The threshold is set against the decay behavior measured in Arjun’s own tests. The content repairs itself on a loop instead of waiting for a quarterly audit. The Semrush AI Visibility Study found that AI citations change 40 to 60% month over month, so a fixed library of any size decays in place without a refresh loop.
Step 6: Track Citations Across All Four Surfaces
The measurement target moves from rankings to citations, mentions, and share of voice, tracked across ChatGPT, Google AI Overviews, Perplexity, and Gemini. AI referrers such as chatgpt.com are segmented in analytics as a distinct traffic class because they convert like referrals rather than like search. AI Growth Agent clients average more than 12,000 additional AI citations and mentions and a 20% or greater lift in impressions across the first twelve weeks, based on AI Growth Agent’s published case study figures, not Arjun’s.
One honest caveat applies to every measurement stack. Buyers frequently copy an AI answer and paste a brand name directly into a browser, which lands in analytics as direct traffic and never gets attributed. Whatever is measured is a floor, not a ceiling. A competitive share of citation for B2B brands in 2025 sits between 5% and 15% aggregate across major AI engines, with 20% or above signaling category leadership.
Step 7: Run a Defensive GEO Audit First
Defensive GEO work protects the brand before growth compounds any mistakes. Before any growth work begins, the audit captures what AI assistants currently say about the business across all four surfaces. A wrong AI answer hurts more than no answer. The visibility audit surfaces hallucinations, competitor misattributions, and outdated claims that need correction before new content compounds on top of a broken record. Defensive GEO runs parallel to growth work on a cycle, because model answers change.
Run a defensive audit on your brand before your competitors do and fix incorrect answers before they harden into the default narrative.
Measurement Stack and Citation Tracking
The measurement stack has three layers that work together. Google Search Console provides the primary signal for impressions, clicks, and decay curves, and the scissors chart where impressions rise while clicks fall gives the first visible evidence of the shift. AI referrer tracking in analytics segments chatgpt.com and equivalents as a distinct source class. Citation monitoring across ChatGPT, Google AI Overviews, Perplexity, and Gemini then tracks share of answer as the headline metric.
Only 38% of citations in Google’s AI Overviews came from pages ranking in the traditional top-10 organic results, across a study analyzing 863,412 unique search queries from October 2024 to February 2025. This 38% overlap underscores why tracking rankings alone produces a dashboard that says everything is fine while revenue stalls. Rankings and citations represent different outputs from different retrieval mechanics.
Wins feed back into production. The system identifies which pages earn citations, doubles down on that structure and language, and queues updates on pages that decay. This feedback loop turns the whole system into a compounding asset instead of a one-time campaign.
Frequently Asked Questions
How long until citations appear?
Citations usually follow a predictable ramp. Coverage and impressions in Google Search Console typically appear within weeks of publishing structured, schema-marked pages aligned to fan-out query language. Citations inside ChatGPT, Google AI Overviews, Perplexity, and Gemini generally follow within one to three months of consistent publishing and refreshing. Compounding, where topical authority accumulates and citation frequency increases, begins after month three. On Arjun’s own site, new articles reached thousands of monthly Google impressions within weeks, and the GEO subfolder became the only source of new impressions on the domain within 60 days. These are his own Search Console figures from his own test lab, not projections.
Can I measure the real impact when buyers copy answers?
Teams can measure a meaningful share of the impact, but not all of it. A significant portion of AI-driven demand lands in analytics as direct or branded search rather than as anything traceable to the AI answer that caused it. The buyer reads the answer, copies a name, and types it into a browser, which shows up as direct traffic. The correct response is to instrument for citations and share of answers across all four surfaces, track AI referrers as a distinct segment, and watch for branded search lift as a correlated signal. Whatever the measurement stack captures is a floor. The real impact is larger, so every measured figure should be treated as a conservative estimate rather than a precise total.
What is the difference between rankings and citations?
Rankings measure position on a human-readable list of results. Citations measure whether a machine-generated answer names, links to, or draws from a specific page when constructing a response to a buyer’s question. The retrieval mechanics are different, the authority model is different, and the success metric is different. SEO earns authority through backlinks and domain tenure. Generative engine optimization earns authority through topical coverage, with structured, fresh, specific pages that answer the questions the machine is actually retrieving against. A page can rank in the top three organic results and still go uncited in AI answers, because the fan-out queries underneath a buyer’s prompt retrieve against a different surface than the keyword the page was optimized for. Conversely, a page outside the organic top ten can earn consistent citations if it is structured correctly, aligned to buyer language, and refreshed on a loop. Rankings and citations represent different outcomes, and optimizing for one does not guarantee the other.
Conclusion: Test, Measure, Repeat
The evidence points in one direction. Buyers have moved their research into AI assistants, citations have become the new vendor-selection mechanism, and the businesses that decode the retrieval layer now will hold positions that compound while competitors pay to catch up. Arjun Karnik’s public test lab documents exactly what earns citations and what does not, with Search Console data, labeled decay curves, controlled fan-out tests, and the misses published alongside the wins. The 7-step system is not a theory. It is the same system producing the visibility on the site where it is documented, which makes the method self-verifying.
Fix the technical plumbing, map the fan-out queries, align to buyer language, publish at machine cadence via AI Growth Agent, install decay tripwires, track citations across all four surfaces, and run the defensive audit first. Test, measure, and repeat, because the game resets weekly and the window for outsized gains is open now.
See the 7-step system applied to your own content library and decide where to start your GEO rollout.
