If your Google rankings are holding steady but your traffic isn’t, you’re not imagining it. You’re watching the middle of the funnel get quietly rerouted through AI search agents and most SMBs have no framework for it yet.
The old goal of SEO was ranking #1 on a results page. The new goal is becoming the answer the source an AI system pulls into its synthesised response, evaluates for trust, and hands to a user who’s already halfway to a buying decision. Search agents like Perplexity and Microsoft’s Bing-powered Copilot don’t return ten blue links. They read multiple pages, cross-check claims, and generate one composite answer with citations attached (or not).
This guide breaks down what’s actually changed, what still works from traditional SEO, and the specific technical and content moves that improve your odds of being cited rather than skipped.
Why Perplexity Deserves Your Attention Now
Perplexity isn’t a niche tool anymore. Query volume has scaled from roughly 230 million monthly queries in August 2024 to an estimated 1.2-1.5 billion queries per month by May 2026, with daily volume in the 35-45 million range. Some more recent industry estimates put the platform past the 1 billion monthly mark outright as it expanded into browser and OS-level integrations.
What matters more than the raw number is who’s asking. Perplexity retrieves and reads candidate pages in real time for every query rather than relying purely on training data, and it cites content published within the last 30 days at a notably higher rate than older content. Visitors arriving from Perplexity have also been shown to convert at a far higher rate than typical organic traffic in independent analyses this is a research-and-decide audience, not a skim-and-bounce one. Traffic here is smaller than Google’s, but disproportionately high-intent. Losing visibility isn’t a rounding error.
Why Bing Is the Hidden Infrastructure of AI Search
Most SEO playbooks still treat Bing as an afterthought. That’s a mistake in 2026. Bing now powers ChatGPT Search, which means ranking in Bing is no longer just about Bing traffic it’s about visibility inside the AI assistant millions of people use every day
Bing’s own share of global search sits around 5% worldwide, according to March 2026 StatCounter data but that number badly understates its reach. Microsoft Copilot is integrated by default in Windows 11 and Edge, meaning Bing benefits from every Copilot search query, and Yahoo Search runs entirely on Bing’s technology. Add in Microsoft 365 Copilot pulling research queries out of Word, Outlook, and Teams, and you have a distribution footprint that reaches far beyond bing.com. For B2B software, professional services, and enterprise content publishers, Copilot is arguably a higher-value AI search channel than raw market share implies.
There’s also a competitive argument here: Bing carries a fraction of the content saturation Google does, generally meaning less competition per query, even as its AI-referral reach keeps climbing. One caveat: an analysis of 41 million results across ChatGPT, AI Overviews, Perplexity, and Copilot found ChatGPT’s cited sources overlap with Bing’s own results by only around 26%, despite ChatGPT’s browsing being Bing-powered so ranking on Bing is a strong signal, not a guarantee of ChatGPT citation.
The Foundation: Traditional SEO Still Matters
Nothing here replaces core SEO it extends it. Independent research is fairly consistent on this point: among the major AI platforms, Perplexity shows the strongest alignment with traditional Google rankings, with one large-scale citation study finding roughly 43.5% URL overlap and 55.2% domain overlap between Perplexity’s citations and Google’s top 10 organic results.
Google AI Overviews shows a similar pattern: SeoClarity’s analysis of 432,000 keywords found 97% of AI Overviews cite at least one source from the top 20 organic results not the top 10, the top 20. So ranking #1 isn’t required, but being nowhere near page one is a real handicap. Crawlability, domain authority, page speed, and clean on-page structure remain the entry ticket nothing below works if Google can’t find and trust your site first.
Technical Optimisation for AI Search Agents
This is the part most SMBs skip entirely, and it’s often the fastest fix.
Audit your robots.txt. Training crawlers (GPTBot, ClaudeBot, Google-Extended) scrape content to train future models, while retrieval bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot, Claude-User) fetch pages in real time to answer a live user question. Blocking retrieval bots removes your content from AI-generated answers on ChatGPT, Claude, and Perplexity entirely and Rutgers/Wharton research found publishers who blocked AI crawlers saw a 23.1% total traffic decline without reliably reducing citation rates. Block training bots if you’re opposed to model training on your content, but explicitly allow the search-facing ones.
Add an llms.txt file. A plain markdown index at your site root pointing AI systems to your best pages. It’s not a ranking factor and adoption is still modest around 10% of sites but it’s nearly free to implement.
Use schema markup deliberately. Bing and its AI ecosystem lean heavily on structured data. Organisation, Person, Article, FAQ, How To, and Product schema all improve extraction accuracy. FAQ schema correlates with roughly 40% higher citation weighting in ChatGPT specifically.
Submit via IndexNow. This notifies Bing (and its AI-powered surfaces) the moment content is published or updated, instead of waiting on a routine crawl.
Write for passage-level extraction. AI agents pull the specific paragraph that answers the query, not your whole page. Every H2/H3 section should stand alone as a complete answer.
Content Strategy for AI Citation
The data on what gets cited is now extensive, and it’s converging on a few consistent patterns.
Format matters, and it varies by platform. Multiple large-scale studies agree listicles, product pages, and homepages are the three dominant content types cited across ChatGPT, Perplexity, and Google AI Overviews, though the mix shifts: ChatGPT leans product-page dominant, AI Overviews lean listicle-dominant, and Perplexity is more balanced between the two. Intent shapes it further all three models cite listicles most overall, but ChatGPT and Google AI favour articles second, while Perplexity uniquely elevates discussion content like Reddit and LinkedIn threads.
Freshness is a real ranking lever. Content published within the last 30 days gets cited by Perplexity at a meaningfully higher rate than older content. A freshness pass on your existing library updated stats, an honest “last updated” date, current examples is often higher-leverage than writing something new.
Answer-first structure beats keyword-stuffed structure. Build sections around the literal questions your audience types. Direct definitions, Q&A formatting, and clear heading hierarchy make it easier for an AI agent to lift a clean, quotable passage.
A real-world note from running this ourselves: when we restructured older AI Outlier articles to lead each section with a direct one- or two-sentence answer before the supporting explanation instead of building up to the point those sections started showing up as extractable passages far more consistently in our own manual Perplexity checks. It’s a small structural change with an outsized payoff, and it costs nothing but an editing pass.
Authority Signals Beyond Your Website
AI citation isn’t purely an on-site game. Google AI Overviews maintains around 54% overlap with traditional organic rankings, but the picture across platforms is far more fragmented research tracking 680 million citations found only about 11% domain overlap between what ChatGPT and Perplexity each choose to cite, meaning strong visibility on one platform doesn’t guarantee visibility on another.
This is where earned media does real work. Brand mentions on reputable third-party sites industry blogs, review platforms, trade publications help models recognise your brand as a credible entity independent of your own domain authority. LinkedIn specifically punches above its weight, consistently ranking among the most-cited domains for professional queries, which makes cross-linking between your site and LinkedIn a genuinely useful, low-cost tactic.
Measuring AI Search Visibility
Traditional analytics weren’t built for this. AI agent referrals frequently get misclassified as “Direct” traffic or don’t register at all, leaving most SMBs flying blind on a channel already sending them qualified visitors.
A few purpose-built tools track this now: Otterly.AI, Profound, and Peec AI all monitor how often a brand is cited across ChatGPT, Gemini, Perplexity, and Bing Copilot a “share of model” metric, roughly the proportion of AI-generated answers in a topic area where your brand shows up.
No budget for a paid tool yet? Start manually. Pick five to ten core topic queries, run them through Perplexity and Bing Chat weekly, and log whether you’re cited, who’s cited instead, and in what format. Not scalable long-term, but genuinely useful and free.
The First-Mover Window Is Open For Now
Most competitors haven’t built for AI citation yet. Robots.txt files are still misconfigured, schema markup is still an afterthought, and content is still written for keyword density instead of passage-level extraction. That gap is exactly why the cost of entry is lower right now than it’s likely to be in twelve months, as more sites catch up and the competition for citation slots intensifies.
Quick-Start Checklist
- Audit robots.txt confirm PerplexityBot, GPTBot, and Bingbot’s search-facing agents aren’t blocked
- Publish an llms.txt file pointing to your best content
- Add Article, FAQ, and HowTo schema to your highest-value pages
- Set up IndexNow for instant Bing indexing on publish/update
- Restructure top pages so each section answers a question in the first sentence
- Refresh publish dates and stats on your most important existing content
- Pursue at least one earned media or LinkedIn cross-linking opportunity this month
- Start a weekly manual citation check in Perplexity and Bing Chat for your core topics
This article reflects publicly available data as of August 2026. AI search platforms update their retrieval and ranking mechanics frequently treat platform-specific percentages as directional rather than fixed, and verify current figures before making major budget decisions.