How is AI changing search? The familiar journey started with a Google query, several open tabs, and repeated comparisons. Now many journeys begin in ChatGPT, Gemini, Claude, Perplexity, or Google AI Overviews. Instead of a list of links, the user receives a synthesized answer with an explanation and a shortlist. This is conversational search.
This changes both the interface and the customer journey. The first impression forms earlier, the shortlist may exist before a website visit, and the click becomes validation rather than discovery. Below we examine how ChatGPT changes customer behavior, the marketing funnel, content, and measurement.
Why it's not "another channel" but a new demand interface
A few public markers to consider:
- Google AI Overviews, since its launch in May 2024, has gradually rolled out to more than 100 countries and has become a standard part of the search experience for a significant portion of queries (Google Blog).
- ChatGPT Search with web grounding launched in October 2024 (OpenAI). OpenAI reported roughly 800 million weekly active ChatGPT users in 2025.
- Back in February 2024, Gartner predicted that by 2026, the volume of traffic from traditional search engines could decrease by about a quarter in favor of AI chatbots and virtual agents (Gartner).
This means that a significant part of the "warm" demand in your commercial topics today begins not in search results, but in a dialogue with AI. The question is not "is it worth considering", but "how exactly".
Old and New User Path
The quickest way to see the difference is to put both routes side by side.
| Stage | Classic Search | AI Dialogue |
|---|---|---|
| Start | Query in Google | Query in ChatGPT / Perplexity / Gemini / Claude |
| First result | List of 10 references | Ready answer with 3–5 options and context |
| Comparison | The user opens 3-5 sites and compares the | It takes place in the dialogue itself, without going to |
| Clarification | New request → new search results | Clarifying prompt in the same chat |
| Creating a shortlist | Based on what the user has read on the sites | Based on what the AI has said |
| Clicking on the site | Often at an early stage — as a way to check | At a later stage — as a check of an already formed choice |
| The role of the brand | Be in the top 10 on request | Be in the collected response as one of the 3-5 options |
| Metrics | Impressions, clicks, CTR, rankings | Mention rate, share of voice, citation rate, and cited sources |
The classic search has not disappeared anywhere. In many niches, it continues to bring in most of the traffic. But the place where a person has a first impression and the first short list of options has already moved.
What a dialogue path actually looks like — with an example
A hypothetical user who chooses a CRM for a small team:
- "What CRMs should you consider for a team of 8 sales in B2B?" — gets a list of 4-5 options and a brief description of each.
- "Which of them is suitable if the team is already using Telegram and Google Sheets?" — the model narrows the list to 2-3 options, adds integration criteria.
- "How much does it cost to implement?" — receives approximate figures for each option.
- "And how is [option A] better than [option B]?" — the model decomposes the difference.
- "Show real cases of implementation in Ukraine" — the model refers to materials in industry media.
- Only after that — click on the website of the selected brand, often directly from the links in the AI response.
Along the way, the user interacts with your brand 5 times — and all 5 before they enter the site. If your brand is not in steps 1-2, you will not appear further — the choice narrows without you.
What's Changing for the Sales Funnel
If you superimpose a dialog path on a classic funnel, you can see a shift:
| Funnel stage | What was here before | What is here now |
|---|---|---|
| Awareness | Saw a banner, brand campaign, or media article | Encountered the brand in an AI answer to an informational query |
| Interest | Searched for general information on Google | Asks the AI to explain the topic and name the players |
| Consideration | Opened multiple sites and compared | Asks the AI to compare options in the same dialogue |
| Intent | Visited the site from a commercial query | Asks AI about prices, case studies, integrations, and risks |
| Conversion | Click → application form → call | Click already on the formed selection or immediately direct contact |
The upper stages of the funnel do not disappear. They just occur more and more often in someone else's interface - in AI responses. If the model in your category doesn't understand how to describe you briefly, you're out of consideration before the sales team notices it.
What this means for content
Content should now close not one request, but a chain of clarifications. Signs of pages that work well in a dialog script:
- explain the topic so that the model can quote a fragment that sounds like an answer;
- have clear H2/H3, lists, tables — a structure from which it is easy to extract short blocks;
- close a series of logically related questions: "what is it → how does it work → how to choose → what is different → how much does it cost → what are the risks";
- contain an FAQ of 5-10 short pairs of "questions → answers";
- have specifics: numbers, dates, examples, cases;
- are updated and have a fresh date.
A separate format that works well in dialogue is comparative materials ("X vs Y", "Top-N for Z"). If competitors in your niche have their own comparisons, and you have only commercial landing pages, you will lose in consideration.
What this means for the brand and positioning
AI does not simply list every well-known company. It tends to mention brands with clear context in public sources. A brand therefore needs more than recognition; it needs a concise, consistent description:
- one category with which you are consistently associated;
- 1-2 clear customer segments ("for teams up to 20 sales", "for e-commerce in Europe");
- 1-2 advantages that are repeated in external descriptions;
- stable reference groups (with whom you are compared in articles and selections).
If it is difficult for a team to quickly answer "who are we for and what is one strong advantage", AI will not be able to either.
What this means for metrics
The old set (positions, clicks, traffic, CTR) remains, but no longer covers the entire path. Metrics are added that describe the presence in the dialog:
- Mention rate — in what proportion of responses the brand appears.
- Share of Voice — how often you are mentioned against competitors.
- Prompt coverage — in which types of requests you are present and in which you are not.
- Model coverage — how many models the brand appears on stably.
- Citation rate — how often models cite your domain.
- Co-citation / adjacency — with whom you are consistently placed.
Details on metrics are in the article AI visibility vs SEO: which metrics are important now.
Why the click now comes later — and what to expect from it
In the classic scenario, the click was an early stage of acquaintance: the user came, reads, forms an opinion. In the dialog, a click often comes with a hypothesis formed by the AI response.
This changes expectations for landing pages:
- a person came to check the formed impression, and not to understand from scratch "what kind of company is this";
- the visitor may already know your offer, prices, and proof points from the AI answer;
- they expect the website to confirm what the AI answer said;
- a contradiction between the answer and the landing page weakens trust;
- the sales conversation starts with a better-informed prospect.
The practical task is clear: the landing page must answer the same questions the user asked the AI and confirm the claims that shaped the visit.
How to adapt: sequence of actions
Work plan, if you're just starting to consider the dialog layer:
- Collect a prompt pool for real decision scenarios: 40-80 category-level, comparison, and problem-oriented prompts.
- Establish a baseline in ChatGPT, Gemini, Claude, and Perplexity. Record where the brand appears, where it does not, which competitors appear nearby, and which sources are cited.
- Break down the user path into 4-6 typical dialogues and see where you fall out in it.
- Analyze sources and competitors in those places where it fails.
- Prepare content for question chains, and not for individual keywords: guides, FAQs, comparisons, cases.
- Finish the site under the "impression check": H1, first screen, FAQ, schema.org, cases with numbers.
- Check in 4-6 weeks: whether the Mention rate, SoV, citation rate have moved.
It's not a new job on top of SEO, it's an extension of it — another layer of the same pages, another layer of the same signals.
Common mistakes in adaptation
- Look at the dialogue as a "channel with traffic". Most of the value here is in the formation of a short list, not in the click.
- Try to "outsmart the model". Prompt injections and hidden text in the model's HTML are either ignored or marked as spam.
- Create a page for every possible prompt. Does not work. High-quality coverage of typical threads of questions works.
- Build everything on the same model. ChatGPT, Gemini, Claude, and Perplexity give different pictures — you need an intersection.
- Expect a quick effect in the training layer. These are months, not weeks.
Frequently Asked Questions
Will AI Dialogue Replace Classic Search? No. Rather, there will be a distribution: information and consideration requests — dialogue; navigational and specific — classic search results. Most businesses will continue to need both.
How to measure that a brand loses at the stage of dialogue, and not later? If there is traffic for brand queries, but you are not in category and comparative AI queries, most likely you fall on the consideration in AI.
Do I need to redesign the main page for AI? Most often, no. Point edits work: H1, the first screen with a direct answer to a key question, adding FAQs, and schema.org. A drastic redesign is usually not necessary.
Is it possible to build a funnel in which AI replaces the sales team? Not yet. AI is good at shaping consideration, but the final contact in B2B and complex products remains human. In consumer categories, a click can go straight to the purchase, but this is a separate scenario.
What about languages? A Ukrainian user in commercial topics in Ukraine usually receives a response in Ukrainian or English, depending on the model and request. Content should be in both languages if you are targeting both local and foreign demand.
What else to read
- What Is AI Visibility And Why Businesses Are Not Enough Anymore With SEO
- AI Visibility vs SEO: What Metrics Are Now Important
- Why your brand doesn't appear in ChatGPT, Gemini, or Claude
- How to Choose the Right Queries for AI Visibility Monitoring
How we do it in VYDAI
The user path in the dialog is difficult to see from GA4 — only the final click is visible there. To reconstruct the dialogue, you need to run typical chains of requests through models and see where exactly the brand falls out.
VYDAI systematizes this process: you create a query set by scenario (category, comparison, problem, and follow-up), and the service runs it through ChatGPT, Gemini, Claude, and Perplexity. It records mentions, citations, and adjacent competitors so you can see where the brand drops out of the journey.
To inspect this journey for your own brand, create an account or view the demo.
