Optimized For Search – Wherever It Happens

Kurt Fischman
Founder, Growth Marshal
Say 👋 On Linkedin!

KURT FISCHMAN | FOUNDER OF GROWTH MARSHAL

Published Mar. 2025

Introduction

Search engine optimization (SEO) is no longer confined to traditional web search engines. While Google still dominates with around 90% of search market share, users are increasingly seeking information through new channels and AI-driven tools.1 Nearly 40% of Gen Z, for instance, prefers searching on TikTok or Instagram over Google’s products.2 Voice assistants, chatbots, and social media platforms have all become search interfaces in their own right. This evolution demands a rethinking of SEO: content must be optimized for search wherever it happens, whether that’s a Google query, an AI-powered answer engine, a TikTok search bar, or a voice query to Alexa.

This white paper explores the technical underpinnings of these emerging search models and provides strategies for ensuring content remains discoverable across them. We delve into algorithmic breakdowns of AI-native search (e.g. ChatGPT and similar large language model systems), AI-augmented traditional search (Google’s and Bing’s new generative search experiences), social media search algorithms (TikTok, Instagram, YouTube), and briefly touch on voice search. Implementation guides and case studies are included to illustrate how SEO practitioners and web engineers can adapt to this multifaceted search landscape. The goal is to equip advanced SEO professionals and developers with a deep understanding of how to optimize content for any search platform.

The Evolution of Search Beyond the Web Browser

For decades, SEO strategy centered on web crawlers and keyword-matching algorithms. Google’s PageRank and later machine-learning updates shaped how content was created and structured. Today, however, search queries might be answered by an AI chatbot synthesizing information from numerous sources, or by a trending video in a social app. Key trends include:

AI-Native Search Engines: Fully AI-driven assistants (like ChatGPT, Bing Chat, Google Bard) that generate answers in natural language instead of showing a list of links.
AI-Augmented Search Results: Traditional search engines integrating AI summaries or chat features alongside or above standard results (e.g. Google’s Search Generative Experience, Bing’s AI answers).
Social Media Search: Platforms like TikTok, Instagram, and YouTube now serve as search engines for many users, especially for local spots, how-to videos, and product discovery. These platforms use their own recommendation algorithms to surface search results.
Voice Search: Voice assistants (Siri, Google Assistant, Alexa) allow users to ask questions verbally. They often read out a single answer, making answer optimization (sometimes called Answer Engine Optimization) crucial.

The diversification of search means SEO can’t focus on a single algorithm. Instead, we must understand multiple algorithms and modalities. In the following sections, we break down how these different search systems work under the hood and what content optimization techniques they demand.

Algorithmic Breakdown: AI-Native Search Models

AI-native search models are systems designed from the ground up to use artificial intelligence (typically large language models) to handle queries. Unlike a traditional search engine, which indexes the web and matches keywords, an AI search model like ChatGPT generates answers from an internal knowledge base or retrieved context. Let’s examine how these models function:

Large Language Models as Answer Engines: ChatGPT (based on GPT-3.5/GPT-4) is a prime example of a generative AI that can answer questions. It is a transformer-based neural network trained on vast swaths of text. Rather than retrieving a document, it produces an answer word-by-word, using probabilities learned during training. The model has encoded knowledge of facts and language patterns from its training data. However, a vanilla LLM has no direct access to current information on the live web; its knowledge may be only as recent as its training cutoff. This is why some AI search implementations combine LLMs with retrieval mechanisms to stay up-to-date and grounded in real sources.

Retrieval-Augmented Generation (RAG): Modern AI search engines often use a hybrid approach: the AI model first retrieves relevant documents from a database or the web, then generates a synthesized answer using those documents as context. In academic terms, these systems are “generative engines” that retrieve and then generate.2 The retrieval step might use a traditional search index or a vector database of embeddings (semantic search), while the generation step uses the LLM to synthesize an answer and possibly cite sources. This paradigm is used by tools like Bing Chat, which performs a web search and feeds the top results into an AI that crafts an answer, and by Google’s SGE (discussed later). It’s also employed by standalone QA systems (e.g. Perplexity.ai or Neeva AI). The key algorithms involved include:

Query understanding: The AI interprets the user’s question, often generating an embedding (a high-dimensional vector) that represents the query’s meaning. Text embeddings are vectors that capture semantic context and relationships between words. This allows the system to go beyond exact keyword matching and find conceptually relevant content.
Document retrieval: Using the query (or its embedding), the system fetches candidate texts. This can be done via keyword search or semantic similarity search. In AI chat search, embedding-based semantic search is common: both queries and documents are transformed into vector embeddings, and nearest-neighbor search finds the closest documents by cosine similarity. This finds relevant information even if wording differs.
Answer generation: The LLM receives the retrieved texts (often with prompts instructing it to use them) and composes a unified answer. The algorithm here is the LLM’s decoding process (e.g. beam search or sampling) guided by its neural network knowledge and the provided evidence. Well-designed systems also inject instructions to encourage the model to quote or cite sources. For example, Bing’s implementation appends each source URL to the text snippet before feeding it to GPT-4, ensuring the model can attribute facts to sources in its response.
Relevance and factuality checks: Some advanced setups include an extra step to verify the answer’s quality. Google’s SGE uses the concept of NORA (“No One Right Answer”) to decide when to invoke the AI – essentially when a question is open-ended or complex enough to need synthesis. Others use rankers or filters to ensure the AI’s output isn’t hallucinated; for instance, an autorater ML program might assess whether the answer properly addresses the query.3

Implications for SEO: AI-native search flips the script on how content is surfaced. Rather than a user scanning many snippets and clicking one, the AI might merge information from multiple sources into one answer. This raises concerns for content creators: your site’s information could be used to answer a question without the user ever visiting your site. However, there are opportunities if the AI does provide citations or if users trust the mentioned sources. Being the one whose content is quoted by the AI confers authority and can lead to referral traffic when citations are present. A recent paradigm called Generative Engine Optimization (GEO) has emerged to address this, focusing on techniques to make content more likely to be picked up by generative answer engines.2

Key Points: To optimize for AI-native search, ensure your content is ingested, understood, and deemed reliable by these models. This means: allow your site to be crawled by AI training datasets and search indexes (no restrictive robots rules for common scrapers), publish accurate and well-sourced information (the AI might use your own citations), and build brand authority. An AI like ChatGPT is more likely to mention known, trusted entities in its answer. We’ll cover concrete optimization steps for LLM-based search in a later section.

Book Strategy Call

Algorithmic Breakdown: AI-Powered Traditional Search

Major search engines like Google and Bing have steadily infused AI and machine learning into their ranking and retrieval algorithms for years. Recently, they’ve also launched explicit AI-powered features (like chat modes and generative summaries) on the search results page. Understanding these changes is crucial:

Google’s AI Integration: Google Search uses AI at multiple stages: query interpretation, result ranking, and result formatting. A milestone was the introduction of RankBrain and later BERT. RankBrain (around 2015) was an algorithm learning to adjust results based on query similarity and user behavior, one of Google’s first uses of deep learning to influence rankings. BERT (2019) is a language model used to parse queries and content in order to understand context and intent. At launch, Google said BERT was impacting 1 in 10 searches by helping better interpret longer, conversational queries10. For example, BERT helps distinguish nuances like “train to Chicago” vs “train from Chicago,” which a purely keyword-based engine might misinterpret. These AI models enable Google to match pages to queries not just by words but by meaning, significantly improving relevance for natural language queries.

Google has also deployed MUM (Multitask Unified Model) for understanding content across languages and even across text/image, and uses machine learning for identifying spam (SpamBrain) and for personalization. Moreover, in 2021 Google began Passage Ranking (sometimes called passage indexing): using an AI model to identify specific passages in a page that answer a query, potentially ranking a page for a query even if the query’s keywords aren’t in the title or meta tags. Passage ranking means that a single relevant paragraph deep in a blog post could be surfaced as a top result if it directly addresses the user’s question. This algorithmic change emphasizes the need to structure content clearly (with headings and logical sections) so that even subtopics can be matched to niche queries.

Most recently, Google is experimenting with the Search Generative Experience (SGE) – an AI summary at the top of search results for certain queries. SGE uses a generative AI (likely based on their Bard/PaLM models) to provide a synthesized answer with cited sources. According to Google, SGE is particularly targeted at NORA queries (No One Right Answer) where the user might benefit from an overview or comparison.3 For example, a query like “best electric vehicle for a family of four” doesn’t have a single factual answer; SGE might produce a paragraph comparing options, drawn from multiple web pages. SGE also links out to the sources it used, in a carousel of cards for further reading. In effect, Google is incorporating RAG directly into the SERP. From an algorithmic perspective, SGE likely uses the top N search results (or some trusted sources) as input to an LLM, generates an answer, and then uses those sources as references for transparency. Google has been cautious with this rollout, running it as an experiment for users who opt in, and limiting it to certain query categories. The presence of SGE means SEOs need to consider how their content might be summarized by Google’s AI. If your page provides a thorough answer to a complex question, it could be part of the AI summary (and ideally, your page link will be one of the citations that users can click).

Bing’s AI Integration: In early 2023, Microsoft’s Bing leapfrogged by integrating OpenAI’s GPT-4 model into its search engine. Bing’s traditional search index remains in play, but now every Bing query can be answered in chat form using the data from Bing’s index. The Bing Chat mode will typically fetch the top results for a query (or even perform multiple searches iteratively) and then have GPT-4 compose a conversational answer, citing the sources. This made Bing’s index relevant in a new way: content that was previously ignored by SEOs (who focused mostly on Google) suddenly mattered if it could appear in Bing and thus in ChatGPT/Bing Chat answers. In fact, OpenAI’s own ChatGPT introduced a web browsing mode that explicitly relied on Bing’s search API. This means if Bing doesn’t index your site, ChatGPT won’t see it either.4 SEOs have taken note: ensuring Bing indexing is up-to-date has become a quick win to gain visibility in AI results. Microsoft has also partnered with companies to integrate plugins and real-time data into Bing Chat (for instance, OpenTable for restaurant reservations), which hints at how AI search might directly interface with services in the future.4

Ranking & Click Patterns in an AI World: One major question is how AI answers will affect user behavior and traffic. Early observations indicate that when an AI snippet fully satisfies the query, click-through rates (CTR) to organic results may drop (similar to what happened with featured snippets). However, when sources are cited, users often do click to verify or read more. We are essentially witnessing the rise of “zero-click searches” on steroids, but with the saving grace that AI chats often encourage exploration of cited links. For SEOs, a key metric will be whether being the cited source in an AI answer drives traffic or brand visibility. Bing Chat, for example, includes numbered citations linking to pages; if your content is excellent and chosen by the AI, you might gain a direct visit from curious users. There is also an aspect of trust: brand mentions in AI answers (even without clicks) could have a marketing value in themselves, increasing brand recognition as an authority.

Case Study: High Google Rankings Fuel AI Mentions

A recent large-scale study (10,000+ queries tested) examined what factors make an AI like ChatGPT mention certain brands in its answers. The findings reinforce that traditional SEO strength translates into AI visibility. Brands that ranked on page 1 of Google had a strong correlation (~0.65) with being mentioned by the LLM, whereas brands with lower search rankings were rarely named. High Bing rankings also showed a positive (though slightly lower) correlation with LLM mentions. Surprisingly, the study found that factors like the number of backlinks or having multimedia content did not significantly influence whether the AI would mention a brand. In other words, the AI’s “knowledge” of prominent websites aligns closely with what ranks well in search – likely because the training data of LLMs and the documents retrieved for RAG are skewed toward those top-ranking, authoritative pages.5

Interpretation: Strong organic SEO creates a virtuous circle: the same content quality and authority that gets you to the top of Google also makes AI systems more likely to consider your site a trusted source. This case suggests that SEO fundamentals (quality content, authority, relevance) are foundational for “AI SEO.” It’s not purely about tricking the AI with new tactics; it’s about being the kind of site that algorithms (search or AI) gravitate to for answers. In practice, this means continuing to build expertise in your content area, earning mentions and references across the web (which train the AI), and ensuring the AI can access your content (proper indexing and maybe even providing structured knowledge that the AI can consume easily).

Algorithmic Breakdown: Social Media Search

Social media platforms have evolved their own search and discovery algorithms, distinct from traditional web search. However, they share the goal of delivering relevant content to users quickly. Today, a user might search within TikTok for a tutorial, or within Instagram for a brand’s page, or on YouTube for product reviews – activities that used to happen on Google. Let’s break down a few major platforms:

TikTok: TikTok’s core algorithm (For You Page recommendation) is famously driven by machine learning analyzing video watch time, engagement (likes, comments, shares), and user interests. Its search function leverages similar signals. When a user enters a search query on TikTok, the app returns a list of videos. The ranking of these videos is influenced by textual matches (the query against video captions, descriptions, and recognized speech or text in the video) as well as engagement metrics and personalization. TikTok is known to automatically transcribe speech in videos and perform OCR on text within videos to understand their content, which means the actual spoken words in a video can make it searchable. Hashtags and captions play a role: including keywords as hashtags or in description improves discoverability for those terms. But unlike Google, TikTok can measure how viewers respond to a video for that query – e.g. do they watch it fully? swipe past it? – and adjust rankings accordingly. In effect, TikTok’s search is a combination of semantic understanding (via text and audio processing) and collaborative filtering (via user behavior data). Another element is freshness and trends: TikTok often surfaces very recent or trending videos for queries, reflecting the fast-moving nature of its content.

One notable insight is that TikTok search often has comparable volume to Google for certain topics. For example, one travel client found TikTok’s monthly searches for a tourist destination were on par with Google’s search volume for the same term, and click-through rates for top TikTok results were significantly higher (often well above 50% CTR) than typical Google CTRs.6 This highlights why “TikTok SEO” is now part of the advanced marketer’s toolkit.

YouTube: YouTube has been the second-largest search engine for a long time (owned by Google). Its search algorithm considers video title, description, tags, and transcript for keyword relevance, but it also heavily weighs user engagement signals (views, likes, watch duration, etc.) and personalization (your subscribed channels, past watch history). An important aspect is YouTube’s use of AI to generate captions (transcripts) and to understand the video content. This means that saying a keyword or topic in the video can help the video appear for that keyword in search, even if the text description is sparse. YouTube also intermixes search with recommendation; after the initial results page, it will suggest related videos. For SEO on YouTube, traditional tactics like keyword research still apply (choose relevant titles and tags), but success is closely tied to creating engaging content that retains viewers. High audience retention and interaction will cause your video to rank higher for searches over time. Additionally, YouTube search results can be filtered by recency, duration, etc., so optimizing metadata (like adding chapters/time-stamps which act like subtopic markers in search) can help.

Instagram: Instagram’s search primarily lets users search for accounts, hashtags, and locations, but recently it also searches post captions and uses AI to identify the content of images. Instagram confirmed that it uses keywords in captions and bios to return results even if a query isn’t a hashtag. The algorithm will try to show the most relevant accounts or posts. It likely factors in popularity and interaction too – for instance, a post that has gone viral under a certain hashtag might rank higher for related keyword searches. However, Instagram search is more limited in that it doesn’t return general web-like content, only what’s within the app’s ecosystem. Businesses optimize by using descriptive usernames, filling profiles with relevant keywords (in a natural way), and using hashtags intelligently. There’s also an emerging practice of adding alt text to images (originally for accessibility) which some believe may be read by Instagram’s search AI for context.

Algorithmic Bias and Walled Gardens: Each social platform is a walled garden – their search results won’t include your website (unless you have a presence on the platform itself). This means traditional SEO (for your own site) doesn’t directly make you visible on TikTok or Instagram. Instead, you need to create native content on these platforms. The algorithms are optimized for user engagement within the app, not for sending traffic out. For example, TikTok search might rank a user-generated video reviewing a product higher than the brand’s official video if the former has higher engagement and watch time for that query. There’s also a social trust component: comments and community interaction may indirectly boost a video’s credibility and thus its ranking.

Social Search as a Google Competitor: A Google executive noted that for certain types of searches (like finding a lunch spot or tutorial), young users often go straight to TikTok or Instagram.1 The appeal is a more visual, immediate answer. Instead of scrolling text results, one can see a quick video demo or review. In response, Google has started incorporating more visual and short-video results for some queries, and even launched features like Google Web Stories. We’re seeing convergence: social platforms are improving search, while search engines are adopting more social-style content.

Case Study: TikTok SEO in Action

A UK-based agency noticed an opportunity in TikTok search for their travel client. They discovered that thousands of users were searching TikTok for travel guides (e.g. “Things to do in [Destination]”), often in numbers comparable to Google searches. The agency implemented a TikTok SEO strategy: they conducted keyword research specific to TikTok (using the app’s search suggestions and third-party tools to gauge search volume for various terms), then produced a series of short videos targeting those queries. Crucially, they optimized each video by speaking the target keywords in the dialogue, adding on-screen text with those keywords, and writing keyword-rich captions. This ensured the TikTok algorithm knew exactly what the video was about. They also leveraged trending music and engaging editing to boost watch time. The result: their videos achieved #1 rankings in TikTok’s search for the target travel keywords, leading to massive view counts. In one case, a TikTok about “Things to do in Newquay” ranked at the top and yielded a click-through rate around 66%, far higher than the typical 35% CTR of a #1 Google result in that niche. This high engagement on TikTok not only drove brand visibility, but also funneled interested viewers down the funnel (some TikTok viewers would subsequently search the brand on Google or visit the website, creating a multi-platform user journey).6

Key takeaways from this case study: Keyword strategy and engagement are both vital on social search. It’s not enough to have the keywords; the content must satisfy users. Also, the interplay between platforms was notable – TikTok success fed into Google traffic later (users seeking more details), exemplifying an omni-channel approach. The case also underlines that SEO professionals should broaden their definition of “search volume” to include platform-specific searches. A savvy SEO now asks: what are the top queries on YouTube in my sector? What are trending searches on TikTok related to my product? The answers might guide content creation on those platforms.

Brief Note on Voice Search

Voice search became a buzzword in SEO a few years ago with predictions that “50% of searches will be voice by 2020.” While such forecasts were likely overhyped, voice queries have indeed grown via mobile assistants and smart speakers. Optimizing for voice overlaps with many of the strategies for AI and traditional search, with a few special considerations:

Conversational Queries: Voice queries tend to be longer and phrased in natural language (“Hey Google, what’s the best sedan under $20k for commuting?”) as opposed to terse typed queries (“best sedan under 20k”). This means content that answers spoken questions directly and in a conversational tone can perform well. Including FAQ-style content on your site – where questions are phrased exactly as a user might say them – can help capture these queries.
Featured Snippets and Quick Answers: Often, voice assistants draw answers from featured snippets or Knowledge Graph results. If your content is the one in the coveted “Position 0” snippet on Google, it’s likely to be read aloud by Google Assistant. Therefore, aiming for featured snippets (by providing concise, direct answers to common questions) is a primary voice SEO tactic.
Structured Data for Voice: Using structured data markup (Schema.org) is particularly helpful for voice search compatibility. Markup like FAQ schema, HowTo schema, and Speakable schema (a markup specifically for news articles to indicate a portion of text suitable to be read aloud) can boost the chances of your content being chosen for voice responses. By clearly structuring content and metadata, you make it easier for the voice AI to parse and determine that your content succinctly answers the query. Implementing structured data “creates clear pathways” for search engines, increasing the likelihood that your content will be picked up as the spoken answer to a user’s question.
Local and Navigation Searches: A large proportion of voice queries are local (“find a nearby coffee shop”) or navigational. Ensuring your local SEO is strong (Google My Business listing, reviews, and relevant local schema on your site) is key to capturing those.
Conciseness and Clarity: Because a voice assistant will typically read out a short answer, content that is verbose may be passed over. Aim to provide a concise answer at the top of your content, with additional details following. This is sometimes called the inverted pyramid style of writing (common in journalism), which proves useful for voice SEO as well. The concise answer can be marked up or highlighted as the summary.

In short, voice search optimization means optimizing for answers. You’re preparing your content to be the one answer read out loud, which is a winner-takes-all scenario. Many of the strategies overlap with AI search optimization, since both are about providing direct answers in context. We now turn to specific strategies and implementation guides that cut across all these new paradigms: AI chat, AI-augmented web search, social search, and voice.

SEO Strategies for an AI-Driven Search Landscape

Adapting to the new search landscape requires both technical and strategic shifts in how we optimize content. Below are key strategies and implementation tips for making content visible and compelling across AI chatbots, generative search results, social platforms, and voice assistants.

1. Structured Data and Semantic Markup

Why it Matters: Structured data (schema markup) helps algorithms understand the content and context of your pages. This has long been true for Google (enabling rich snippets, knowledge panels, etc.), but it’s even more critical now as AI systems rely on structured information to generate answers. When content is clearly tagged (e.g., this text is a recipe’s ingredients list, or this paragraph is a definition of a term), an AI can more easily incorporate it or attribute it. Structured data is also used to feed knowledge graphs which power succinct answers and voice responses.

Implementations:

Add Schema.org markup relevant to your content type: articles, recipes, FAQs, products, videos, local business info, etc. For instance, FAQPage schema on a FAQ section can make you eligible for featured snippet answers and is particularly useful for voice search (the assistant can read the Q&A).
Use Article schema and include the author, datePublished, etc., which can build trust. For AI, clearly indicating the author and credentials (e.g., a medical article by a doctor) could boost the content’s perceived reliability.
Implement Organization schema linking to your social profiles, and sameAs links to Wikipedia/Wikidata if your brand or entity has a page there. This helps connect your site to the broader knowledge graph. If an AI chatbot is aware of your entity (e.g., your company is recognized as an entity with certain attributes), it may be more likely to include it in answers.
Mark up definitions or key facts with appropriate schema (e.g., use <dfn> HTML for definitions along with schema annotations). If your page defines a term, you want Google’s and OpenAI’s algorithms to know that’s a definitive description.
For voice, implement Speakable schema for news articles: you can wrap a short summary intended to be spoken. While this is niche, it signals to Google Assistant exactly what to read.

Best Practices: Keep the structured data accurate and consistent with the page content (don’t try to stuff hidden info). Use Google’s Rich Results Test and Bing’s Markup Validator to ensure the syntax is correct. Also, monitor in Google Search Console’s enhancements reports to see if your structured data is being picked up. Remember that structured data not only boosts chances of special features, but it generally aids AI understanding. As one voice search guide put it, this markup “makes it easier for search engines to lift your content and repurpose it verbally or on a smart device” as answers.

2. Content Optimization for Retrieval (RAG readiness)

When AI systems use retrieval-augmented generation, they pull in excerpts from content to feed into the model. Optimizing for this means making your content easy to retrieve and quote. Key approaches:

Use Clear Headings and Segmentation: Break content into logical sections with descriptive headings (H2, H3, etc.). A retrieval algorithm may use headings to identify which section of a page to pull. For example, if a page has a section titled “How to change a tire,” that section is more likely to be retrieved for a query “How do I change a tire?” than a page where the info is buried in a wall of text. Google’s passage indexing is an example of this – a clear section can rank even if the page title isn’t an exact match.
Optimize Passages for Standalone Sense: Each paragraph or section that might serve as an answer should be somewhat self-contained. If an LLM selects 3-4 sentences from your page, will those sentences make sense on their own? It’s wise to write content in a way that important points are complete and attributable even if taken out of full context. This also increases the chance that a featured snippet or voice answer grabs your text.
Include Facts, Figures, and Quotes: Empirical data and succinct statements in your content can be enticing for an AI to use. A study found that content with quotes and statistics earned significantly more visibility in LLM-generated answers, presumably because the AI likes to back up its statements. By providing data points (and even citing sources yourself), you make your content more “generatable”. For example, an AI answer to “benefits of solar energy” might favor a blog that says “Solar can reduce household electricity bills by 50%” with a citation over one that just says “Solar can reduce bills.” The specificity and presence of a citation (even if the user doesn’t see it) can influence the AI.7
Leverage Embeddings (Advanced): Some forward-thinking organizations provide their own content via embeddings or APIs for AI consumption. For instance, a company might publish an open API or a dataset with their knowledge base. While not mainstream SEO, this is an emerging tactic: essentially feeding the AI your content in a structured way. Another approach is maintaining a presence on platforms that AI models draw from – e.g., having your content summarized on Wikipedia or cited in widely used data sources. This bleeds into digital PR, but with the aim of getting into the AI’s training data or preferred knowledge set.

In summary, think about how a smart assistant would extract info from your page. The easier you make that extraction (through structure, clarity, and authoritative info), the more likely your content will be chosen by the AI algorithms in the retrieval step.

3. Entity-Based SEO

As search engines and AI shift from keywords to concepts, structuring your SEO around entities (people, places, things, ideas that are recognized as distinct entities) is crucial. Entity-based SEO is about aligning your content with the knowledge graph and the topics the algorithms understand.

Key Tactics:

Identify Your Key Entities: Figure out the primary entities relevant to your content. These could be your brand (as an organization), your products, or the topics you cover. Use tools or Google’s NLP API to see what entities are detected on your pages. Ensure those are indeed the entities you want to be associated with.
Create Content Hubs Around Entities: Rather than isolated blog posts targeting single keywords, develop comprehensive content around the entities/topics. For example, if your site is about nutrition, recognize “Vitamin D” as an entity and have content that fully explores it (benefits, sources, recommended intake, related concepts like calcium absorption). This depth signals to search engines that your site is an authority on the entity “Vitamin D.” In entity SEO, topic coverage and depth often trump exact keyword repetition.
Link Entities in Content: Use internal linking and contextual mentions to connect related entities. If you have a page about “Eiffel Tower,” mention related entities like “Paris,” “Gustave Eiffel,” “Champ de Mars,” etc., possibly linking to pages on those. This mimics the interconnected nature of a knowledge graph and helps algorithms see the relationships. Google’s Hummingbird and RankBrain updates enabled it to better understand these relationships; e.g., knowing “Eiffel Tower” is a landmark in Paris and linking to “Paris attractions” content can boost the relevance.9
Get into the Knowledge Graph: One of the most impactful (but challenging) steps is to have your entity be recognized in Google’s Knowledge Graph. Typical ways to do this include having a Wikipedia page or Wikidata entry for the entity, being mentioned on authoritative sites, and using schema markup (as mentioned earlier). When your entity is in the Knowledge Graph, Google can directly show info about it (like a Knowledge Panel), and AI systems will have an easier time validating facts about it. For example, if you run a notable podcast, creating a Wikipedia page that the podcast is an entity could result in voice assistants readily answering “Who hosts [Your Podcast]?” by drawing from that data.8
Monitor and Optimize Entity Sentiment: This is a newer idea – AI and search algorithms don’t just note that an entity is mentioned, but also context and sentiment. Ensure your brand is mentioned in positive or authoritative contexts online. For instance, many mentions on forums like “I had a bad experience with BrandX” could potentially color the AI’s training data. While this isn’t fully under your control, brand reputation management overlaps with SEO here.

The shift to entity SEO was well summarized by Google’s move to treat “things not strings.” By optimizing for entities, you align your strategy with how modern search understands the world. In practice, SEOs are finding that a combination of schema markup, content depth, and digital PR (to get those entity mentions on third-party sites) pays off in both traditional and AI-driven search. If the AI knows who/what you are, it’s more likely to include you in answers. As a bonus, entity-focused content is naturally more evergreen and robust against algorithm changes than thin, keyword-focused pages.

4. Optimization for Social Search Platforms

To capture searches happening within social apps, you must apply SEO thinking to your social content. This is more about content strategy and social media optimization but targeted at search behavior on those platforms.

TikTok & Instagram:

Perform keyword research within the app. Use TikTok’s search bar autocomplete to see popular queries. Note trending hashtags or challenges relevant to your domain.
Treat the video description and on-screen text as your “meta data.” Include the key query phrase naturally in the first line of the description. If feasible, have the presenter in the video actually say the phrase early on (because TikTok transcribes audio). In one case, saying the exact phrase “things to do in [City]” in the video helped that video rank for that search.
Use relevant hashtags, but don’t overstuff. 3-5 well-chosen hashtags (mix of broad and specific) can help categorization. For Instagram, hashtags are still one of the main ways posts get into search results. On TikTok, hashtags matter but the algorithm also looks at caption text and even the content itself.
Engage quickly: The algorithms consider how much engagement a video gets soon after posting. Promoting your video to get likes/comments (perhaps sharing it on other channels) can boost its ranking in search results for its keywords. This is akin to a “freshness boost.”
Use platform-specific features: For example, TikTok allows you to add a location (which could make you show up in location-based searches) and now has a text search ad feature where you can ensure your video appears for certain searches via paid promotion. Stay updated on such features.
Analytics and Iteration: Use TikTok’s and Instagram’s analytics to see how people found your content. TikTok’s analytics can show the percentage of views that came from search. This is a vital feedback loop – if certain videos unexpectedly get a lot of search traffic for some term, lean into that topic or format more.

YouTube:

Follow classic video SEO: keyword in title (within 70 characters for display), a detailed description (at least 250 words, incorporating synonyms and related terms naturally), and tags (tags are less important than they once were, but still add a few relevant ones).
Custom thumbnails: While not directly an algorithm factor, an enticing thumbnail will increase CTR when your video shows up in search. Higher CTR can lead to higher ranking (YouTube’s algorithm observes satisfaction metrics).
Encourage engagement and watch time: Ask viewers to like or comment (which can indirectly help ranking). Structure your video content to minimize early drop-off – hook the viewer in the first 15 seconds. High average watch duration signals YouTube that your video is a good result for the query.
Transcripts & Closed Captions: Uploading an accurate caption file or using YouTube’s auto-caption (and then editing any mistakes) can improve the video’s discoverability. The text in the transcript is used for search and also for the “auto chapters” feature which can appear in search results now (Google sometimes shows a specific video segment for a query).
Playlist and Channel SEO: Organize related videos into playlists named after key topics. Playlists can rank in YouTube search too. Ensure your channel about page contains relevant keywords and links (if someone searches your brand on YouTube, a well-optimized channel page should appear).

Cross-Platform Synergy: Use social media to bolster web SEO and vice versa. For example, a popular YouTube tutorial might be embedded in a blog post that can rank on Google – allowing you to capture traffic both on YouTube and on Google (and the blog can link back to your site for conversions). Or a TikTok video that goes viral could be added to a page on your site as rich content, improving that page’s engagement signals. We are moving toward holistic search optimization, where the idea is to meet the user wherever they search and ensure consistent messaging and presence.

5. Content Quality, E-E-A-T and AI

It’s important to state that despite all the new techniques, the core principle remains: create high-quality, useful content. Google’s guidelines emphasize E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness). AI systems, trained on vast data, often implicitly pick up on these signals too. For example, an AI might “know” which sites are considered authorities in a domain by how often they are cited or the tone of text around them in training data.

Strategies to bolster E-E-A-T which directly or indirectly help across search platforms:

Have clear author bylines with author bios that state qualifications (for content in YMYL – Your Money Your Life – categories, this is vital).
Get reviews and mentions on authoritative third-party sites. A mention in a major news outlet or a high-quality backlink doesn’t just help Google rank you higher; it also becomes part of the AI training data and knowledge network surrounding your brand.
Maintain content and update it. AI might latch onto outdated info if that’s what it finds. Keeping pages fresh (and indicating last updated dates) can signal that your info is current, which is good for both crawling and user trust.
Monitor for misinformation and correct it. If an AI gives a wrong answer about your brand (for example, “Brand X was founded in 2010” when it was 2015), address this by making sure the correct info is clearly stated on your site and maybe on public knowledge bases. We’ve seen instances of companies having to “SEO their way” out of an AI’s incorrect association by publishing rebuttals or clarifications.
Avoid clickbait/generic content: With generative AI, there will be an avalanche of mediocre, auto-generated content on the web. Standing out with unique insights, original research, or genuinely helpful guidance will become even more of a competitive edge. AI or not, users (and thus algorithms) will gravitate to content that actually satisfies their query. High bounce rates or poor user feedback will harm visibility across platforms (e.g., if users keep saying “Alexa, that wasn’t helpful,” or if they immediately scroll past your video, those signals are noted).

Conclusion

The expansion of search beyond the traditional Google box is a disruptive change for SEO practitioners, but it’s also an opportunity. By understanding the algorithms driving AI chatbots, generative search features, social media discovery, and voice assistants, we can adapt our optimization strategies to ensure content remains visible and influential. Technical SEO now must account for machine learning factors: from how an LLM might interpret a sentence, to how a video algorithm measures engagement.

In summary, winning “wherever search happens” means:

Covering your bases (structured data, mobile friendliness, fast performance, all still count in every arena),
Speaking the language of AI (using entities, clean structure, and factual, referenceable content),
Embracing new platforms (treating TikTok or YouTube not as afterthoughts but as integral parts of search strategy),
Continuously learning. The AI search field is rapidly evolving – e.g., new GPT models, algorithm updates, or platform policy changes could shift best practices. SEO professionals should stay up-to-date with research and experiments (as we saw with the case studies) and be ready to pivot techniques.

Ultimately, the heart of SEO remains understanding what users are looking for and delivering it in the best possible way. The mediums have multiplied and the answers might be spoken by a computer or shown as a video, but the principle of optimizing for the user’s search experience is constant. By applying the strategies outlined and keeping a flexible, informed approach, we can optimize content for search – wherever it happens.

References

Samantha Delouya, Business Insider: "Nearly half of Gen Z is using TikTok and Instagram for search instead of Google..." (July 13, 2022)
Pranjal Aggarwal et al., Generative Engine Optimization (GEO), arXiv preprint arXiv:2311.09735 (Nov 2023)
Google Blog: “AI updates: Bard and new AI features in Search” – Google’s explanation of Search Generative Experience and NORA queries (May 2023)
Jessica Bowman, Search Engine Land: "ChatGPT Search makes Microsoft Bing an SEO priority" (Nov 5, 2024)
Nick Haigler & Christina Blake, Seer Interactive Study: "What Drives Brand Mentions in AI Answers?" (Jan 7, 2025)
Carrie Rose, Rise at Seven Blog: "TikTok SEO case study: Ranking #1 for a client in the Travel space" (Feb 27, 2024)
Morningscore.io Blog: “LLM Optimization explained (LLMO) – How to optimize for AI search” (2023)
Ryan Sylvestre, NoGood.io Blog: "Schema for Voice Search: Structured Data for Voice SEO" (Oct 15, 2024)
PageOptimizer Pro Blog: "What is Entity-Based SEO and How Does it Work?" (2023)
Google Blog: "Understanding searches better than ever before" – Google’s announcement of BERT in Search (Oct 2019)

Kurt Fischman is the founder of Growth Marshal and is an authority on organic lead generation and startup growth strategy. Say 👋 on Linkedin!