Key takeaways
AI Overviews now appear on more than 48% of Google searches, up from about 6.5% a year ago. Getting cited inside an AI answer matters as much as ranking position.
ChatGPT Search reportedly cites only around 15% of the pages it retrieves, and front-loaded, self-contained answers are the strongest predictor of getting picked.
FAQ and HowTo rich results are gone from Google Search entirely as of May 2026, though FAQPage schema still helps AI systems parse and cite Q&A content.
Blocking the wrong AI bot can quietly cut a site out of AI citations: 71% of top news sites block AI retrieval bots, including PerplexityBot, alongside training bots.
58% of consumers have used voice search for local business information, so Google Business Profile accuracy and LocalBusiness schema still carry real weight.
Google reports that AI Overviews now appear on more than 48% of its search queries, up from about 6.5% a year earlier. Voice search has kept pace, with estimates putting regular use at around one in five internet users. Search is no longer a list of ten blue links. It is a mix of spoken queries, AI-generated summaries, and conversational answers from tools like AI Overviews, ChatGPT, Gemini, and Perplexity.
Two disciplines have grown out of this shift:
Generative engine optimisation (GEO): makes content easy for AI systems to find, understand, and cite in tools like AI Overviews, ChatGPT, and Perplexity.
Answer engine optimisation (AEO): structures content so it can be lifted directly into featured snippets, voice answers, and AI-generated responses.
Both sit alongside traditional SEO rather than replacing it. Here's what actually works right now.
Write for how people actually ask questions
Voice and AI queries are longer and more conversational than typed searches. Instead of "best SEO tools," someone might ask, "What are the best tools for voice search SEO?"
To match that pattern:
Target long-tail, question-phrased keywords, such as "how to optimise for voice search" or "what is AI search optimisation."
Write in a natural, conversational tone, as if answering the question directly.
Use question-based subheadings, followed immediately by a concise 40- to 60-word answer.
Follow the direct answer with supporting detail, an example, or a short list.
This format does double duty. It is what Google looks for when selecting featured snippets and is close to the "answer capsule" structure that recent research found ChatGPT favours when deciding what to cite. ChatGPT's search mode reportedly cites only about 15% of the pages it retrieves. Front-loaded, self-contained answers are consistently the strongest predictor of getting picked.
Use schema markup, but know what it's actually for
Schema markup helps search engines and AI systems understand what a page is about and who is behind it. It is less useful than before for winning visible search features but still matters for how AI systems parse and cite content.
Worth prioritising:
Article or BlogPosting schema with author, datePublished, and dateModified fields, since freshness and named authorship both factor into which sources AI Overviews and Google's newer AI Mode choose to cite.
Organisation and Person schema, to establish a clear entity behind the content.
LocalBusiness schema for any location-based content.
A reality check on two schema types still often recommended: Google restricted FAQ rich results to a small set of government and health sites in 2023, then removed them from Search entirely on 7 May 2026. HowTo rich results were deprecated earlier, also in 2023. Neither type produces a visible search result anymore.
FAQPage schema is still worth using when the content genuinely is a Q&A, because ChatGPT Search, Perplexity, and AI Overviews all parse it as structured, citable content, but treat it as an AI-parsing tool now, not as a way to win a rich snippet.
Make sure AI crawlers can actually reach your content
This is easy to overlook, and it undercuts everything else on this list if it's wrong. Most major AI platforms now run two separate bots: one for training their models, one for live retrieval and citations. Blocking the wrong one can quietly remove a site from AI search results.
OpenAI runs GPTBot for model training and OAI-SearchBot for ChatGPT Search citations.
Anthropic runs ClaudeBot for training and Claude-SearchBot for Claude's search citations.
Perplexity retrieves and cites content in real time through PerplexityBot.
Google keeps Gemini and AI training under Google-Extended, separate from standard Googlebot indexing.
That distinction plays out messily in practice. Among top news sites, 79% block at least one AI training bot via robots.txt, yet 71% also block AI retrieval bots. PerplexityBot, used for indexing, is blocked by 67%. This means many publishers shut out the exact citation traffic that opting out of training was never meant to cost. Check robots.txt to confirm retrieval and citation bots are allowed, even if training bots are blocked. It is possible to opt out of AI training while staying fully eligible to be cited in AI answers. That is the combination most brands seeking AI visibility now choose.
Search intent still decides everything
Every page should match the intent behind the query it's targeting:
Informational queries ("how does AI change SEO") need a clear explanation, ideally backed by a definition and supporting schema.
Transactional queries ("voice SEO services for local businesses") need comparisons, proof points, and a clear next step.
Navigational queries ("[brand] AI search optimisation") come down to entity clarity: consistent naming, Organisation schema, and internal linking.
Get the technical and local basics right
Most voice searches happen on mobile, so performance and structure still carry real weight:
Strong Core Web Vitals, since speed and stability affect both rankings and AI Overview eligibility.
HTTPS and clean, crawlable markup.
Accurate, complete Google Business Profile listings for any location-based business. Incomplete profiles are far less likely to surface for "near me" queries.
LocalBusiness schema and genuine customer reviews, which AI systems use as a trust and relevance signal.
Local intent and voice search overlap heavily. BrightLocal's 2018 survey found that 58% of consumers used voice search to find local business information in the previous year. This figure is still widely quoted because local "near me" behaviour has not changed much since. For the fundamentals behind that, including Google Business Profile setup, reviews, and location-based schema, see the local SEO guide.
Track what's actually working
AnswerThePublic, AlsoAsked, and Keyword Insights for discovering conversational query patterns.
Frase and Surfer SEO for structure and clarity during optimisation.
Google Search Console and GA4 for traditional performance measurement.
InLinks and Kalicube Pro for entity signals, plus a dedicated AI-citation tracker to see how often a brand actually gets mentioned inside ChatGPT, Perplexity, and AI Overviews.
SEO, voice search, and GEO at a glance
Traditional SEO | Voice search optimisation | GEO | |
Optimises for | Ranking position | Spoken, conversational queries | Being cited inside an AI-generated answer |
Content shape | Keyword-led pages | Question-and-answer format | Fact-dense, self-contained passages |
Key signal | Backlinks, on-page SEO | Featured snippet eligibility | Entity clarity, freshness, and crawler access |
Getting cited in an AI-generated answer is not about chasing a single tactic. It comes down to clear structure, accurate schema, crawlable AI bots, and genuine expertise on the page. These are the same fundamentals that have always mattered, just measured differently now. Not sure where your site currently stands? Our SEO services start with exactly that kind of audit.
FAQs
What is voice search optimisation?
Structuring content so voice assistants and search engines can lift it directly into a spoken or written answer: natural language, clear structure, self-contained responses.
How is AI changing SEO in 2026?
AI Overviews now appear in more than 48% of Google searches, and platforms like ChatGPT and Perplexity generate answers rather than just linking to results. Ranking well still matters, but getting cited inside those answers now matters just as much.
Do I still need FAQ or HowTo schema?
Neither produces a visible rich result in Google Search anymore; both were phased out between 2023 and 2026. FAQPage schema is still worth using for genuine Q&A content, since AI systems still parse it; don't expect a snippet from it.
How can a local business improve its voice search visibility?
Keep Google Business Profile details accurate, add LocalBusiness schema, and collect genuine reviews. BrightLocal found 58% of consumers had used voice search for local business information, driven largely by complete, accurate listings.
Do AI crawlers need special permission to access a site?
Most AI platforms respect robots.txt but run separate bots for training and for live search citations. Make sure citation bots like OAI-SearchBot, Claude-SearchBot, and PerplexityBot are allowed, even if training bots are blocked.
