Model changes are now dropping overnight with zero warning, and AI search engines are fundamentally changing how users discover brands.
For this edition of Sip & Search at BrightonSEO, we brought together Aleyda Solis, International SEO Expert and Founder of Orainti; Nick Lafferty, Founding Marketing Engineer at Profound; and Tom Capper, Senior Search Scientist at Moz, to map out how to secure and protect visibility in Large Language Models (LLMs).
In this recap, you will get direct expert insights on how LLMs pull data, how to optimise for fragmented search intent, and the specific playbooks required to earn valuable third-party citations.
AI search engine volatility and query fan-out
1. Model changes drop constantly and affect citation patterns. How should brands build a strategy that can survive platform shifts?
Nick Lafferty, Founding Marketing Engineer, Profound
Every brand needs a dedicated monitoring tool because AI Engine Optimisation (AEO) is fundamentally different from traditional SEO.
In traditional search, professionals are accustomed to receiving updates from the Google Search Liaison stating an algorithm update is rolling out over two weeks. This warning system does not exist in AI search; platforms will drop a new model overnight without warning.
When your executive team demands to know how a sudden model shift impacts your business visibility, having no data because tracking was paused is a worst-case scenario.
To build a strategy that withstands platform shifts, you must consistently monitor your visibility over a long time horizon.
2. Moz found that 88% of citations in Google’s AI Mode don’t appear in the organic top 10 for the exact same query. How can brands improve topic coverage to optimise for query fan-out?
Tom Capper, Senior Search Scientist, Moz
Source: Moz
To understand that statistic, you must look at how AI Mode functions. It presents its output to the user as a single search, but it is actually executing multiple diverse searches in the background.
The queries it runs internally are so different from what the user originally typed that the chances of the same websites ranking organically are heavily marginalised.
Because of this, relying entirely on your website’s first-party content will not cut it. As SEOs, we have been conditioned since 2018 to believe that writing better content on our own domains is the definitive answer.
In the era of AI search, your strategy must explicitly include websites other than your own. It is highly unlikely you will hit the necessary topic and intent coverage if your entire strategy is limited to your own domain.
3. What does the agentic commerce ecosystem look like right now, and what should brands be doing to prepare?
Aleyda Solis, International SEO Expert & Founder, Orainti
The ecosystem is in its testing infancy. While OpenAI was the first mover with ChatGPT Shopping Research, early integration results showed suboptimal conversion rates compared to a brand’s standard online store.
E-commerce sites are fully optimised to upsell, cross-sell, and prevent cart abandonment; brands lose control of the conversion experience if they hand it off without backend data or transparent analytics.
Because of this, many e-commerce and retail brands are sceptical. They prefer to let competitors make the initial mistakes. Conversely, Google already has this infrastructure built out via Google Merchant Centre. There is a much higher chance of commercial success in Google’s AI Mode because Gemini is halfway there, with structured data and Merchant Centre integrations. OpenAI was the first mover, but they will likely not be the ultimate winners.
4. Traditional tracking lists fail because valuable user prompts almost always have a search volume of one. How is prompt tracking different, and where should a brand start?
Tom Capper
The first thing to understand is what not to do. Do not simply copy your existing keyword list, paste it into a prompt tracking platform, and consider the job done. Nobody types traditional search keywords into an LLM.
When analysing prompt research data, queries with significant search volume are usually conversational phrases like “thanks, try again,” which offer no business value. The prompts that actually offer commercial value almost always have a search volume of one.
To make this data meaningful, queries must be aggregated into broader topics. When you are building example prompts to fill those topical buckets from scratch, your immediate cheat code is People Also Ask data. You already have access to that long-tail query data, and it is an excellent starting point.
How AI models retrieve content and citations
5. Content depth and sentence count positively correlate with AI citations, while backlinks show negative correlations. How do you square that with traditional SEO?
Sources: Growth Memo – What Content Works Well in LLMs, Profound – AI Search Shift Research
Nick Lafferty
Good performance in AI search still rests on solid SEO fundamentals. In the early days of LLMs, backlinks and traffic did not impact how often you got cited. As models evolve and dynamically pull in content, such as ChatGPT using APIs to search Google, backlinks and traffic do matter. Ultimately, the best thing you can do is write highly effective content for humans, not just for AI agents.
However, winning a citation requires a different approach than ranking on page one of Google.
For example, when ChatGPT evaluates a webpage, it does not always ingest the whole page. It focuses heavily on three tiny elements:
- The snippet: A summary snippet of roughly 150 characters.
- The page title: Clear, direct semantic labelling.
- The URL structure: Logical entity hierarchy.
6. AI search still feels very English-first. What should global brands do to improve visibility across languages and regions?
Aleyda Solis
Some marketers are questioning whether they should stop investing in localised content assets, given that LLMs rely heavily on English-language content for citations. That is a mistake.
These models were initially trained on English-first data, but they are getting significantly better with every release. We cannot build long-term marketing plans based on a temporary weakness of the discovery platforms.
Instead, look for gaps and understand which local third-party assets make sense for your specific region. The prominent digital PR sources or social communities in the US will not be the same in a market like Brazil.
You must research the top local media outlets, local communities, and local ccTLDs that are ranking. This requires digital PR and community management teams that thoroughly understand their local ecosystems, as answers will increasingly be sourced locally over time.
7. With monthly citation drift hovering between 40% and 60% across major platforms, how can brands distinguish between a genuine visibility drop and normal AI volatility?
Source: Profound
Nick Lafferty
It comes down to looking at your data over a broader time horizon. Do not focus on whether your visibility dropped yesterday or today. Instead, look at the trend over a rolling 7- to 14-day period and account for seasonal factors like regional holidays.
Because an LLM can give a completely different answer every time it is prompted, you must look at a longer data horizon to separate normal algorithmic volatility from a true drop in performance.
8. Robots.txt and blocking AI crawlers: is it a strategic competitive move or self-sabotage?
Juliette van Rooyen, audience member
I previously worked with a large e-commerce group where we publicly blocked many bots. We could do that because we were selling products, not information.
If you are selling information, you absolutely need rules to ensure that users consume it on your site. But if you are an e-commerce site selling physical products, you lose nothing by providing that data freely.
All you do by blocking bots is tell the AI agent you don’t exist, preventing it from actively promoting your products for consideration.
Oleg Korolov, audience member
I broadly agree, but there are edge cases. Major media institutions like the Financial Times block robots to protect their writers’ work, and they can afford to do so because they are a definitive, irreplaceable source. Unless you are running a true institution with no real competitors, blocking bots is risky.
Philippine Sikora, audience member
It is also worth remembering that robots.txt is largely an educational guideline. Bots do not always follow the rules; many will continue to crawl your site whether you want them to or not.
Measuring AI visibility and ROI
9. We have a dark funnel problem where B2B buyers form vendor lists via AI conversations before ever contacting sales. How do you turn this influence into measurable ROI?
Source: Similarweb
Aleyda Solis
The biggest mistake we make today is treating traffic and clicks as the ultimate measure of AI presence, or directly correlating them with on-site conversions. That is the floor, because many attribution paths are currently unmeasurable via traditional frameworks.
I recommend building a staged assessment and measurement framework:
- Track presence first: Audit where your brand is surfaced or missing across critical AI user journeys to find content gaps.
- Internal proxy metrics: Use qualitative user data collection, such as post-purchase surveys, to identify hidden channels.
- Third-party proxy metrics: Monitor tools like SimilarWeb or Semrush to map your relative AI traffic market share against competitors.
Nick Lafferty
If you are in B2B, you should absolutely ask a required, open-ended “How did you hear about us?” question on your sales form. Marketers often push back, claiming conversion rates will tank if you add that field. My view is that if someone refuses to fill out your form because you asked how they found you, they weren’t a high-quality lead anyway.
We use this at Profound and defend it religiously. HubSpot might tell me that only 5% of our leads come from LinkedIn, but when I look at the open-ended form responses, the number jumps to 30%. This qualitative data directly shapes how we invest our business capital.
10. AI systems often cite the clearest answer, not the most complete page. What does AI-optimised content look like without making the page feel thin or robotic?
Tom Capper
My biggest advice here is not to overthink it. A lot of current advice advocates over-gaming the way you structure your content. That is a short-term play at best, and terrible for your human users at worst.
Your answers need to be clear and consistent, and critically, they must exist beyond your own website.
Aleyda Solis
This ties directly back to how lean teams can secure a third-party presence through community management and digital PR.
Traditionally, large brands keep these departments in organisational silos that ignore SEO. AI search makes the value of breaking those silos tangible because we are dealing with citations rather than backlinks. Tools can now showcase citations at the domain level, making it clear when competitors are recommended over us because of user-generated content, news mentions, or YouTube reviews.
It is the citation that matters now, and that citation must align with your product’s unique selling proposition and values.
Tom Capper
Whatever the answer is to the most important question your brand answers, don’t just host it on your website. Populate it across other digital spaces where your brand can influence the wider discussion.
11. AI search engines most often cite third-party content. How can lean marketing teams get more third-party citations if they don’t have a budget for digital PR?
Laura Lancu, audience member
If you want to be cited and trusted on a budget, you must focus on community building. Platforms like Reddit and Quora are completely free and readily available channels to reach your audience.
Judith Lewis, audience member
If you have no budget but have time during operational slow periods, identify free, publicly available databases relevant to your industry. Conduct meta-research by pulling different data points together into one comprehensive, high-value research piece. You can then use the unique data asset to pitch individual media outlets during your downtime.
90-day action plan for LLM visibility
12. If a brand had only 90 days to improve AI visibility, what should they prioritise first?
Tom Capper
Focus heavily on off-site content and traditional branded SERP optimisation, but look beyond just Google.
Optimise how you show up on YouTube, LinkedIn, Wikipedia, and Medium. It is astonishing how heavily AI engines, particularly Gemini, cite YouTube content.
Ensure your core brand answers are managed consistently across these third-party platforms rather than obsessing over on-page text structures on your own site.
Aleyda Solis
Identify immediate media opportunities to pitch your existing data. This allows you to piggyback on big publications that are already trusted sources for LLMs without having to build links from scratch.
Nick Lafferty
Focus entirely off-page, but start by tracking prompts first.
Use a platform to figure out exactly which domains are currently being cited for your specific industry vertical.
Marketers often read that Reddit or Wikipedia are the top-cited domains globally, but citation patterns are highly industry-dependent. If Reddit isn’t being cited in your specific space, don’t waste time there. Gather the data for your niche first, and let that dictate your off-page strategy.
Conclusion: The future of visibility is integrated
To win visibility in the era of LLMs, brands must completely dismantle internal marketing silos. AI search engines do not look at SEO in isolation; they evaluate your brand’s entire digital footprint, rewarding clear content structures, consistent off-site messaging, and strong multi-channel integration.
Johnson Ishola
Johnson Ayomide Ishola is a content marketer at the FCDC. With a background in Economics, he brings a strategic, analytical edge to storytelling and brand development.
He specializes in lead generation and community-driven content, focusing on building authentic audience connections. At the FCDC, Johnson develops initiatives across various events and programs, using data-backed insights to craft narratives that drive engagement and growth.
