The era of the "answer": how to optimise content for AI search and overcome the challenge of zero clicks

There was a time, not so long ago, when our relationship with search engines was a simple list of blue links. We would type a query (often awkwardly, like "best pizza restaurant Rome centre") and Google, in return, would offer us a list of possibilities. We would click, read, go back, and click again. We were the ones doing the work of synthesis.

That time is ending.

Today, when we ask a question, we expect an answer. Not a list of links, but a clear, concise, and reliable summary. Generative artificial intelligence, integrated directly into search interfaces (think of Google's "AI Overviews", Perplexity, or ChatGPT-powered searches), has changed the rules of the game.

This is the new reality of content marketing. Organic traffic is no longer guaranteed, even if you are in the top spot. AI, acting as an "assistant" that reads the entire web on the user's behalf, often generates a complete answer that makes a click unnecessary. It is the dreaded era of "zero-click searches".

Many see it as an apocalypse for SEO. We, at Insight, see it as an evolution.

It is no longer just about "Search Engine Optimisation" (SEO). It is about "Answer Engine Optimisation" (AEO), or if you prefer, "Generative Engine Optimisation" (GEO). The goal is no longer just to rank, but to become the authoritative source that the artificial intelligence chooses to cite.

If AI is a synthesiser, our job is to provide it with the best and clearest ingredients. This in-depth guide is not a list of "tricks", but a strategic and technical manual for restructuring how you create content, so that both AI and users will choose you.

Understanding the "brain" of AI: why E-E-A-T is your best ally

To win this game, we must stop thinking of Google as a simple crawler that indexes keywords and counts backlinks. The large language models (LLMs) that power these new searches "reason" differently.

The technical fundamentals still count, but the algorithm now tries to mimic human judgement. And how does a human being assess quality? Through trustworthiness.

This is where E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) comes in.

Whereas E-E-A-T was previously a "soft" concept for assessing page quality, it is now a technical signal that LLMs actively seek out. It is no longer enough to say you are an expert; you must prove it in ways that an AI can interpret.

Experience

AI is trained to distrust generic content that looks like it could have been written by anyone. It looks for evidence of first-hand experience.

How to demonstrate it:

• First-person content: Use "I" or "we" as a team. Tell a story
• Real case studies: Instead of "5 ways to improve X", write "How we improved X for client Y (with data)"
• Original photos and videos: Stop using only stock images. If you are talking about design, show your desk. If you are talking about a product, show it in use. Original photos and screenshots are powerful signals of direct experience

Expertise

Who is speaking? Why should we trust this person? AI asks itself these questions.

How to demonstrate it:

• Clear author bios: Every article must have a recognisable author. Not a generic "staff"
• Detailed author page: The bio must link to a comprehensive page listing the author's credentials, publications, professional social media (e.g. LinkedIn), and their background
• Thematic consistency: The author (and the site) must consistently write about the same topic. A blog that discusses finance, then recipes, and then gardening will have a very low perceived expertise by AI

Authoritativeness

It is not just about how many links you receive, but who is talking about you.

How to demonstrate it:

• Quality backlinks: Links from thematically relevant and authoritative sites in your industry
• Mentions (even without links): AI scans the web looking for "co-occurrences". If your brand is often mentioned alongside terms like "marketing strategies" on other sites, podcasts, or forums, the AI learns the association
• Original data: Publish research, surveys, or infographics with proprietary data. Become the primary source that others cite

Trustworthiness

Does the user (and the AI) trust browsing your site? Do they trust your information?

How to demonstrate it:

• Transparency: Clear and easy-to-find "About Us", "Contact" (with a physical address and phone number, if applicable), and "Privacy Policy" pages
• Sources and citations: If you report data or facts, cite your sources. Show you have done your research
• Security: HTTPS is the bare minimum, but it is a fundamental signal of trustworthiness
• Reviews and testimonials: Social signals that validate your company or brand

Ignoring E-E-A-T in 2025 is like building a house without foundations. The AI simply will not deem it "worthy" of being used as a source for its answers.

Answer architecture: how to format for AI extraction

Imagine your article not as a single piece of text, but as a database of answers.

Artificial intelligence, when it needs to generate a summary, does not "read" your article from beginning to end like a novel. It "scans" it, looking for clear, modular, and easy-to-extract blocks of information. Your job is to format the content so the AI can grab the right "Lego brick" at the right time.

If your article is one dense block of text, the AI will ignore it. If it is a set of well-organised "bricks", it will use it.

The "gateways" for AI: featured snippets

Before AI Overviews existed, there were featured snippets (the "position zero" boxes with the direct answer). These are not dead; on the contrary, they have become the main training ground and the preferred source for AI answers.

Optimising for featured snippets is the first step to optimising for AI.

How to do it:

• Identify a clear question (e.g. "How do you prune olive trees?")
• Provide a direct, concise, and precise answer (40-60 words) immediately below the heading
• Use bulleted or numbered lists for processes ("step 1", "step 2"...)

Semantic structure: the instruction manual for bots

HTML code is not just for "styling" text. It is the instruction manual that tells the AI the hierarchy of the information. Using HTML tags semantically is fundamental.

Essential tags:

• H1 (Heading 1): Only one per page. It is the main question or theme
• H2 (Heading 2): Use them for the main questions that make up the article. Think of them as chapters in a book
• H3 (Heading 3): Use them for sub-questions or to detail the points of an H2
• Lists (<ul>, <ol>): AI loves lists. They are structured, easy to analyse, and perfect for being repurposed in a summary answer
• Bold (<strong>): Use it to highlight key concepts and entities (names, places, brands) within a paragraph
• Short paragraphs: 2-3 lines are ideal. This makes extraction easier

The "sandwich method" for the perfect answer

This is one of the most effective techniques. For each H2 or H3 (which should be a question), structure the content like this:

[H2] What is multimodal SEO?

[Direct answer - The "bread"] Multimodal SEO is the optimisation of not only textual content, but also visual, audio, and video content, to answer user searches that use complex queries (e.g. voice + images).

[Further detail - The "filling"] Unlike traditional SEO, which focuses on keywords, multimodal SEO considers how search engines (like Google Gemini) interpret the combined context of images, text, and audio...

[Example - The other "bread"] For example, optimising a product image with descriptive ALT text and accompanying text that explains its use allows the AI to answer a question like "find me comfortable red shoes for running".

This format is perfect for AI. It has the short, citable answer (the first paragraph) and the in-depth context immediately after.

Structured data (Schema Markup): precision labelling

If semantic HTML is the instruction manual, structured data (Schema.org) is the set of precision labels.

It is a technical job that requires inserting JSON-LD code into the <head> of the page. But it is also the most powerful way to explicitly tell Google and the AI: "Hey, this piece of text isn't just text, it's an FAQ" or "This isn't just a list, it's a recipe".

The most important schemas for AI optimisation:

• FAQPage: If you have a question and answer section, you must mark it up with this schema. It is gold dust for AIs looking for answers to specific questions
• HowTo: If your article is a step-by-step tutorial, this schema formats it perfectly for extraction
• Article: Specifies the author, publication date, and modification date. Crucial signals of freshness and E-E-A-T
• Person: Used on the author page, it specifies who that person is, their social profiles, and their qualifications

This "modular" and structured approach not only helps AI, but it also drastically improves readability and the user experience. Users, like AI, scan text looking for answers. A clear structure serves both.

Beyond words: multimodal optimisation and thematic authority

New-generation AI models (like Google's Gemini or GPT-5o) are natively multimodal. This means they do not just "read" text; they "see" images, "listen" to audio, and "watch" videos.

Ignoring multimedia content means leaving half of the optimisation opportunities on the table.

Images as "proof" and "context"

The era of generic, purely decorative stock images is over. Every image must have a purpose and must add informational value.

What to use:

• Annotated screenshots: If you are explaining software or a process, screenshots with arrows, circles, and notes are fundamental. The AI reads text within images
• Original diagrams and infographics: A complex concept explained with a clear diagram is a powerful signal of expertise. (As a creative studio, we use generative AI to create these unique diagrams, not to simulate stock photos)
• Real photos (Experience): Photos of the team, the product in use, or "behind the scenes"

Technical optimisation (ALT Text):

The ALT tag has become the main textual descriptor that AI uses to understand an image.

• ❌ "Bad" ALT: alt="article-photo"
• ⚠️ "Okay" ALT: alt="dog running"
• ✅ "Excellent" ALT (for AI): alt="Diagram illustrating how optimising for featured snippets is the first step for AI optimisation"

File name: The file name also counts.

• ❌ No: DSC8012.jpg
• ✅ Yes: content-optimisation-ai-search-checklist.jpg

Building thematic authority (Citable Brand)

The AI must learn that your brand is synonymous with a certain topic. This goes beyond individual articles and concerns your long-term brand strategy.

The goal is to become a "citable" brand.

How to build it:

Co-occurrence and mentions: AI learns from associations. If you want to be the authority for "drone videomaking", you need to ensure your brand is mentioned (even without links) on drone podcasts, videomaking forums, and in technology reviews.

Become the source: The quickest way to be cited by AI is to publish original data. A survey on your industry, a market analysis, a case study with real numbers. AI is hungry for fresh data and statistics. If you are the only one with that data, the AI will be "forced" to cite you.

The network of authority: Link your content. If you have a pillar article on "photography", make sure all your smaller articles (e.g. "portrait technique", "lens choice") link to that pillar article. This creates an internal "network" that signals to the AI which content is the most important on your site for that topic.

The technical "customs": Robots.txt and crawling control

You can have the best content in the world, but if the AI does not have permission to read it, it is all useless.

The Robots.txt file

This file, located in the root of your site (yoursite.com/robots.txt), is the "bouncer" that tells bots (crawlers) which pages they can visit and which they cannot.

The risk: Many sites, fearing that AI will "steal" their content to train its models, have started blocking AI bots (like OpenAI's GPTBot or Google-Extended, which Google uses for Gemini's training).

The problem: If you block Google-Extended or other bots that feed AI answers, you will never be able to appear in AI Overviews.

The strategy: You must find a balance. It is wise to block access to irrelevant pages (e.g. test areas, thank-you pages, internal archives). But you must absolutely ensure that your blog articles, service pages, and pillar content are accessible to AI bots.

Check your robots.txt now.

The (new) LLMs.txt file

This is still an experimental file, but it is gaining traction. It is similar to robots.txt, but it is designed specifically to give instructions to large language models (LLMs).

While robots.txt is a "yes/no" (you can enter / you cannot enter), LLMs.txt is more granular and mainly serves to manage copyright and use for training.

Adopting it now, even if experimental, shows that you are a technically advanced publisher and aware of the new dynamics.

From "content creators" to "answer architects"

The shift from a search for links to a search for answers is the biggest shake-up that SEO and content marketing have seen in the last decade.

It can be frightening. Seeing organic traffic eroded by an AI box that summarises our hard work is frustrating. But complaining is not a strategy.

The opportunity is enormous. AI, by its nature, needs sources. It prefers structured, authoritative, up-to-date content written by real experts. In one fell swoop, this evolution is sweeping away years of "rubbish" content (content fluff) and keyword stuffing.

It is rewarding those who create real value.

Our role as editors, marketers, and creatives is changing. We are no longer just "content creators"; we are becoming "answer architects". Our job is to build informational structures so solid, well-designed, and reliable that artificial intelligence cannot help but use them as foundations for its answers, citing us and recognising our authority.