Original insights by Aleyda Solis
30-second rundown
Key learnings:
- AI Search influence escapes normal traffic reports: Customers can discover, compare, and validate a brand inside an AI platform without clicking, so referral sessions reveal only the minimum impact.
- Topical coverage must support buying decisions: Create content that answers customer questions at every stage: discovering the problem, comparing options, making a purchase, and getting help afterwards. Keep important facts in crawlable HTML.
- Third-party evidence shapes brand authority: Reviews, coverage, links, communities, and comparison pages help AI systems verify how a company should be described and recommended.
This week: Choose one profitable product, group five prompts by customer journey stage, and check whether your brand appears, is accurately described, and receives a linked citation. Fix the weakest page by adding a concise answer, explicit decision criteria, and crawlable product facts.
Coming 3 months: Use an AI Search monitoring tool to repeat approved prompt groups daily and record mentions, citations, sentiment, accuracy, competitors, and landing pages. Assign one owner to review monthly changes beside branded search, assisted conversions, survey responses, leads, and revenue.
Aleyda Solis has opened the hood on AI Search and found the old SEO engine still running, only now it is hauling a casino, a courtroom, and a robot salesman through the desert. Her argument is not to burn the playbook. It is to expand what SEO measures and influences before LLMs control the shortlist.
Traffic remains useful, but it can no longer testify alone. AI systems can introduce a brand, compare it with rivals, validate its claims, and influence a purchase before analytics records a visit.
1. The Click Is Now the Minimum Evidence
AI referral traffic is a baseline, not the full impact. Solis recommends separating business outcomes from visibility signals instead of pouring everything into one fraudulent “AI impact” number.
Core outcomes include revenue and purchases from LLM referrals, AI conversion rate, and AI-assisted conversions. Visibility metrics include share of voice, mentions, citations, sentiment, prompt coverage, recommendation rate, linked citation rate, comparative win rate, and representation accuracy.
Evidence has three levels. Observed data is recorded directly, such as a visit or sale from an AI referral. Proxy data is indirect evidence that AI influenced someone, such as branded searches, direct traffic, social mentions, or customers saying AI introduced them to the company.
Modeled estimates use these signals and assumptions to estimate revenue influenced by AI Search. They help with planning, but are not proof. An estimate can wear a clean suit, but remains an estimate under the interrogation lamp.
Track stable prompt groups because AI answers vary. Organize them by product, audience, journey stage, market, and commercial priority. A running-shoe retailer might track wet-weather shoes for commuters separately from low-cost trail shoes, then compare recurring patterns across relevant platforms.
2. Build Pages That Survive Retrieval
Mapping a pile of queries to one landing page is cracking apart. Query fan-out lets an AI system expand one prompt into related searches, while synthesis can pull useful passages from different pages. The goal is complete decision-stage coverage with several pages, not one swollen page impersonating an entire business.
Cover awareness with guides and research, consideration with comparisons and reviews, decision with pricing, availability, demonstrations, case studies, and compliance information, then support customers with documentation and troubleshooting. Start where demand or AI visibility exists, then expand into adjacent questions.
Decision constraints are the details that change a recommendation. “Best running shoes” becomes a different hunt when the buyer adds flat feet, £100 budget, London rain, or road use. State product attributes consistently, supported by relevant structured data, comparisons, filters, buying guides, and internal links.
Each important section should stand alone with an answer-first summary and descriptive heading. Tables and bullets make comparisons easier to extract. Expert content still needs current evidence, clear entities, technical accessibility, and enough context to remain accurate when lifted from the page.
Do not hide critical facts or links behind client-side JavaScript. AI crawlers may not render it like Googlebot. Put essential information in initial HTML or reliable server-side rendering, and keep media crawlable with alt text and relevant surrounding copy.
Use crawler testing, server logs, and technical validation to confirm that important pages, feeds, links, and media remain accessible to search engines and AI bots. This foundation also prepares transactional sites for agentic commerce, where an AI agent needs dependable product details before comparing an option.
3. Your Website Cannot Corroborate Its Own Story
Solis cites AirOps research finding that 85% of brand mentions in AI Search came from third-party sources, while warning that the percentage will vary. The practical message is the loaded revolver on the desk: your website is only one witness.
Relevant backlinks, digital PR, independent reviews, communities, videos, directories, and comparison sites can confirm or contradict a brand’s positioning. Context matters. A mention is not useful if it is negative, inaccurate, outdated, irrelevant, or attached to the wrong differentiator.
Run an AI Search gap analysis across priority prompts. Record which sources recur, which third-party pages support competitors, how LLMs describe your brand, whether outsiders repeat your differentiators, and where inaccurate descriptions persist. Then assign each gap to SEO, digital PR, community work, customer service, or product messaging.
Participation must be genuine. The objective is not to manufacture applause in forums like a crooked casino manager stuffing the audience. It is to earn accurate representation in the places buyers already trust.
SEO, PR, social, brand, product, and community work all feed the same public evidence layer. Give one owner responsibility for priority claims, their external proof, and errors needing correction.
AI Search does not abolish SEO. It expands the battlefield from rankings and clicks to retrieval, citations, sentiment, accuracy, third-party corroboration, and commercial outcomes. Experiment, keep confidence labels honest, and change course when the evidence changes.
The dashboard may never record every customer AI systems influenced. But a company that measures the right clues, publishes usable answers, and earns credible outside validation can see where the tire tracks disappear into the desert.
