generative engine Optimisation just got more complicated. New research confirms that ChatGPT’s Thinking mode and Instant mode pull from different source pools, with only a quarter of cited URLs overlapping. For our clients, this means a single AI visibility strategy is no longer enough. We’re now treating each reasoning mode as a distinct channel — and adjusting our content, video, and technical SEO workflows accordingly.

Key Takeaways

• Only 25% of sources cited by ChatGPT overlap between its Thinking and Instant reasoning modes, demanding mode-specific optimisation strategies.
• Video SEO now directly feeds AI citation pipelines — not just YouTube rankings and Google video carousels.
• URL parameter hygiene is a quiet but critical factor in how AI crawlers parse and trust your pages.
• The “agentic web” is here: AI agents are making purchasing and recommendation decisions on behalf of users, and brands need a structured optimisation stack to stay visible.
• Competitor keyword analysis must now span traditional search, AI Search, and web conversations to close real visibility gaps.

ChatGPT’s Two Modes Cite Different Brands — Plan for Both

This is the headline finding. Research published by Semrush reveals that only 25% of cited sources overlap between ChatGPT’s different reasoning modes. Thinking mode tends to favour in-depth, research-heavy content. Instant mode leans toward concise, authoritative pages. Same prompt, different brand recommendations.

We’ve started auditing client content libraries against both modes. The practical step: produce two tiers of content per core topic. A deep-dive piece targeting Thinking mode citations. A tight, fact-dense page built for Instant mode. Our early tests show improved citation frequency within weeks.

Video SEO Now Feeds AI Visibility, Not Just YouTube

Video content is no longer just a YouTube play. Google’s AI Overviews and ChatGPT both pull from video transcripts, metadata, and structured markup. We’re following the latest guidance on how to optimise video for YouTube, Google, and AI citations to ensure our clients’ video assets surface across all three channels.

Key actions we’re taking:

• Adding full transcripts with keyword-rich timestamps to every client video.
• Implementing VideoObject schema on all embedded video pages.
• Writing video descriptions that answer specific conversational queries AI models favour.

URL Parameters: The Technical Debt AI Crawlers Won’t Forgive

Messy URL parameters create duplicate content signals, waste crawl budget, and confuse AI indexing systems. We’ve integrated parameter audits into our standard client maintenance workflows, aligning with updated technical guidance on how URL parameters work and when to use them. Canonical tags, parameter handling in Google Search Console, and clean URL structures are non-negotiable basics that directly affect whether AI systems trust and cite a page.

Optimising for the Agentic Web: AI Agents Are Choosing Brands for Users

AI agents — autonomous systems that browse, compare, and recommend on behalf of users — are already influencing purchase decisions. This is not theoretical. Brands that fail to structure their data for agent consumption will simply not appear in agent-driven recommendations.

We’re implementing the five-layer optimisation stack outlined in Semrush’s guide on how to optimise for the agentic web. That means structured data, machine-readable pricing, clear product specifications, authoritative backlink profiles, and consistent entity information across every platform an agent might query.

Competitor Keyword Gaps Now Span Three Surfaces

Traditional keyword gap analysis only covers organic search. That’s incomplete. We now run competitor audits across traditional search, AI-generated answers, and web conversations such as Reddit and forum threads, following the methodology detailed in Semrush’s resource on finding competitor keywords and closing visibility gaps. The brands winning in 2025 are the ones visible wherever their audience asks questions — whether that’s Google, ChatGPT, or a niche subreddit.

The GEO landscape has fractured. A single-channel optimisation strategy is a liability. Our team is treating each AI reasoning mode, each content format, and each discovery surface as its own optimisation target. That’s the baseline now.

Frequently Asked Questions

What is Generative Engine Optimisation (GEO) and why does it matter in 2025?

GEO is the practice of optimising content so it gets cited and recommended by AI systems like ChatGPT and Google’s AI Overviews. It matters because AI-generated answers are rapidly replacing traditional search results as the first point of brand discovery.

How do web designers ensure a site is visible across ChatGPT’s different reasoning modes?

You need both in-depth, research-backed content for Thinking mode and concise, authoritative pages for Instant mode. Auditing your content against both modes and filling gaps is the most direct way to improve citation rates.

Why does video SEO affect AI search visibility?

AI models parse video transcripts, metadata, and schema markup when generating answers. Properly optimised video content can earn citations in AI Overviews and ChatGPT responses, not just YouTube rankings.

What is the agentic web and how should brands prepare for it?

The agentic web refers to AI agents that autonomously browse, compare, and recommend products or services on behalf of users. Brands need structured data, machine-readable content, and consistent entity information to be selected by these agents.