If you’ve experimented with an AI chatbot to answer a question rather than using a classic search engine, you’ll be aware of several things: 1) Sometimes AI gives you a quick answer, sometimes it doesn’t; 2) When it does, it’s a huge time-saver when you don’t have to plough through a lot of search page links; and 3) if it doesn’t work the first time, it’s easy to refine your question and have another go. 

AI-powered search promises more tailored questions and answers and that’s going to impact search referrals:

• SparkToro found that only 36% of searches in 2024 generated clicks to the open Web.
• Seer Interactive research shows click-through rates fell by 70% when Google displays an AI overview.
• Orbit Media polling suggests half of Americans expect AI to eventually replace search engines.

For those worried about the impact of AI on Search, here’s a guide to what’s going on and how to prepare for a new era.   

1. AI-generated content is crowding out the good stuff

The history of the Google algorithm has been a battle on two fronts. The first is a cat-and-mouse fight over screening out bad actors – those whose intent is to game the system rather than provide quality content.

Google’s long-standing battle against low-quality content has entered a new phase. While the “Helpful Content” updates of 2023 and 2024 aimed to reward originality, AI’s ability to mass-produce blog posts has overwhelmed Google’s filters. Publishers still report AI-generated content outranking more authoritative sources, suggesting Google is struggling to enforce its own rules — at least for now.

2. Users are impatient and want Answer Engines not Search Engines

On the demand side, searchers are falling out of love with Google. Search results pages are increasingly cluttered with ads and spammy content, whereas users want rapid answers. Most don’t click links on mobile search results pages and AI-enhanced search is the top ask for consumers, according to one survey. 

People aren’t abandoning Google, but they are frustrated. Google’s response? The rollout of Search Generative Experience (SGE) offering AI-generated summaries right on the results page. This change aligns with user expectations: fewer clicks, faster answers.

3. AI is better at understanding what searchers want

The second battle has been to uncover searchers’ true intent. When Google started you just got a search results page with 10 links on it. And Google did that by matching keywords in content to those in queries. There were no ads. 

Nowadays, Google uses AI to guess what searchers are really looking for and you now get a mini webpage with links, ‘knowledge graph’ content like Wikipedia summaries, pictures, lots of advertisements, and much more. 

Increasingly, you get an ‘Answer Box’ with a straight answer to your query in a featured excerpt.  Google doesn’t share data on how frequently this happens, but research from SEOClarity suggests 10-20% of total queries.  

But there’s a limit to what you can achieve with short queries in a search box. 

The to and fro of an interactive conversation with a GenAI chatbot suggests that, eventually, GenAI tools will be dramatically better at this. Sophisticated users are already using GenAI tools to suggest the questions that should be used for an effective prompt – ‘tell me what I need to tell you to answer my question fully’.

4. The Doomsday Scenario is that Search goes the same way as Social

Once the source of huge volumes of traffic to websites, Social Networks have progressively deprioritised content with links back to other sites. The Content Marketing Institute estimates social networks now generate just 2% of site referrals against 64% for Search. 

It’s not clear how traffic is going to be affected but technology and consulting firm Gartner has added to the anxiety by predicting that search referrals will fall by 25% by 2026, while a model from Search Engine Land predicted falls of 18-64% for 23 websites as a result of Google’s Search Generative Experience.

Google is reported to want to support search referrals to publishers and advertisers. Microsoft has said it is trying to use AI to boost traffic to websites.

But Bill Gates is on record as saying that GenAI and the dawn of personal digital assistants could mean people no longer need to visit search engines again (or Amazon or lots of other current digital necessities).

Survey evidence from Search Engine Journal suggests 40% of millennials are willing to make the switch to GenAI search tools.

But a combination of inertia and the big Search engines’ evolution mean that evidence is scant that any wholesale behaviour change is happening.

ChatGPT has only been available to the wider public since November 2022 and only modest shifts can be traced so far. Bing, which beat Google to integrating GenAI into search, has barely increased market share, according to StatCounter GlobalStats. Meanwhile, Google’s market share has fallen from a little over 92% to a little under 90%. This looks like it will be a slow-motion revolution.

5. There’s an issue with AI’s lack of transparency

It’s hard to understand how GenAI tools reach their conclusions. This makes it hard to debug errors or ensure fairness. ChatGPT will give you an answer, but unless prompted will give you no sources, and even then is vague.

Google made behind-the-scenes efforts to be fair when AI image creation was made available on Gemini, but the launch ended in ridicule after the company admitted it had restricted the ability to create images with white people.

6. Hallucinations add to AI’s ‘trust gap’

GenAI tools can just make things up. References to non-existent academic papers and news articles have been one of the problems.

The FTC issued warnings in 2024 about ‘synthetic facts,’ and benchmarks like Vectara’s Hallucination Leaderboard show major differences in model reliability.

Google Search might include links to unreliable sites, and the answer box may not give you the answer you need, but it’s not editorialising content for you – the final decision on what’s relevant and accurate is up to you. 

As long as AI tools occasionally get it wrong, users will look for signals of credibility — like links, named authors, and cited sources.

7. The tools are evolving fast and hybrids are aiming to build trust

The line between search engine and chatbot is blurring. Microsoft’s Copilot enhanced Bing and Google’s SGE are converging towards tools that answer questions, cite sources, and generate content on demand. Meanwhile, platforms like Perplexity and YOU AI combine clarity, links, and follow-ups to redefine the search experience.

• Perplexity is perhaps the leading challenger at the moment. At first sight, It looks like the google search box page. But when you type in a question, it generates an answer and highlights the sources used. It also suggests further questions. You can restrict searches to particular sites, like YouTube for instruction videos, or academic material.
• Andi is the most visually arresting of the new arrivals. There’s a particular focus on useful links, and It offers a tool to summarise those links for those too busy to read. Controversially, you can view articles within the tool rather than having to go to publishers’ sites.
• YOU AI majors on digging deeper with suggestions for further questions and a ‘research’ mode that provides a particularly rich set of links. It also has a ‘create’ mode that helps turn answers into potential content.

All these engines have free versions and there are no ads. While they are built for convenience, they are strikingly generous in providing links to the underlying sources they have used. 

Additionally, the Answer Engines seem to be making a much better job than Google of weeding out sites that are derivative and spammy. 

That gives rise to a more optimistic scenario in which transparency and convenience mean that links remain as crucial to the Answer Engine Experience as for Search Engines. 

8. Reasons to be cheerful: The Expertise Premium

It is highly likely that overall search referrals to websites will fall – the sheer convenience of answer engines will drive that. But speed is only one factor, confidence in results is just as important. This is where AI is likely to struggle and will need to borrow authority from sites with genuine human expertise.

Initial research by Semrush on Google’s Search Generative Experience indicates that analysis and opinion material is favoured by the model. Likewise authority in niche topics. 

9. How Website Owners Should React

This isn’t about abandoning SEO. It’s about evolving it. Here’s how:

• Double down on originality: AI detectors are unreliable, but answer engines are getting better at spotting generic content. Resist the temptation to churn out derivative content using AI and try to be original and human.
• Optimise for conversations: You should be doing this anyway given that 20% of search queries are now via voice. AI requires refocusing on natural language questions rather than keywords and entities. 
• Use schema markup: Structured data makes it easier for engines to parse and summarise your content.
• Build author authority: Add bios, credentials, and experience to every byline.
• Diversify your channels: The threat of reduced Google search referrals raises the value of other channels. Social networks are increasingly used like search engines with YouTube the second biggest after Google. And subscription-based channels like newsletters and instant messaging have much higher engagement rates than you’ll ever get on Search or Social. Treat newsletters, YouTube, and LinkedIn as core traffic sources, not side projects.