AI Search Optimisation - HiveMCR Masterclass

AI Search Optimisation HiveMCR Masterclass

Schedule THE AUTHENTIC SEO COMPANY | AI SEARCH MASTERCLASS |
HIVEMCR Session 1: 10am – 12pm Intro: AI Search Is a Layer, Not a Channel Chapter 1: AI Search: What Is It & Acronyms Chapter 2: AI Search Process - RARER Methodology Chapter 3: What is the Retrieval Layer? The Technical Stuff Lunch:12pm -1pm Session 2: 1pm – 3pm Chapter 4: AI Search: Content Strategy Chapter 5: Tips & Tools Chapter 6: Reporting & The Future 3 - 3.30pm – Break 3.30 - 4.30pm – Trainer Presentation to MMU Students 7pm – HiveMCR Opening Social at Impossible

Intro AI Search Is a Layer, Not a Channel

AI Search Is a Layer, Not a Channel THE AUTHENTIC
SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR Layer 1 The Traditional Search Layer Google, Bing and other search engines. Where most commercial intent and conversion still takes place Layer 2 The AI Retrieval Layer AI overviews, AI mode, ChatGPT, Claude, Perplexity, Gemini. Synthesis and citation, rather than ranking

Two Search “Lenses” - One Organic Strategy THE AUTHENTIC SEO
COMPANY | AI SEARCH MASTERCLASS | HIVEMCR SEO Lens Classic Keyword Universe Targeting • Search Volume and Classification • Search Intent Analysis • Clustering, Pillars, Spoke & Hub Content • Rank Tracking (still important for a range of reasons even if there is no true “rank”) AI Search Lens Semantic Coverage • Entity & Concept Depth • Fan Out Query Analysis • Passage Level Extractability • Prompt Level Citation Tracking (flawed, but potentially useful) • Topic Level Visibility (more useful)

Today’s Big Takeaway - A Holistic Organic Search Strategy THE
AUTHENTIC SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR One Holistic Strategy

Chapter 1 What Is It & Acronyms

THE AUTHENTIC SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR

There Are Already Too Many Acronyms THE AUTHENTIC SEO COMPANY
| AI SEARCH MASTERCLASS | HIVEMCR GEO (Generative Engine Optimisation) A strategy designed to make content easily understood and surfaced only by generative AI tools like ChatGPT. Unlike traditional SEO, which prioritises ranking URLs through content quality and perceived authority, GEO focuses on optimising for clarity and structure so snippets can be referenced and featured in AI-generated responses. This approach primarily applies to LLMs. AEO (Answer Engine Optimisation) A marketing strategy that focuses on optimising content to be directly featured in answer boxes, voice assistants and AI chatbots. The goal is to provide quick, concise answers to user questions without requiring them to click through to a website. LLMO (Large Language Model Optimisation) This approach optimises for visibility within AI generated summaries and answers. It focuses on semantic relevance, clear answers and conversational language rather than just keywords. AIR SEO (AI Ready SEO) A comprehensive SEO strategy designed to rank your site across all surfaces. This covers technical SEO, content strategy, offsite SEO, UX and includes optimising for AI search (LLMs).

We’ll Chat About Entities Later, But They’re Not Yet a “Thing”

You Can’t Just Create Terminology & Expect It To Rank

It’s Time For a Change of Mindset THE AUTHENTIC SEO
COMPANY | AI SEARCH MASTERCLASS | HIVEMCR Instead of asking yourself “How do we drive more organic clicks or sessions” (rankings, traffic, sessions) We now also need to consider “How do we get answers extracted from our content” (LLM visibility, citations, extractability and brands winning without clicks) This is a fundamental shift, but one that will impact SEO and AI Search, as Google starts to bake this method of retrieval into most, if not all, of its ranking algorithms. But this is an evolution of organic search, not a new channel

There’s No Doubt We’re In the Post AI Era of
Search THE AUTHENTIC SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR Present Day The Post AI Era • AI Powered Analysis • Retrieval Augmented Generation • Multi Modal Understanding The Semantic Era • Semantic Search • Hummingbird Algo • Entity Recognition & Knowledge Graphs • Personalisation & Content Awareness 1990s - 2010s Foundations of Search (Traditional Rankings) • Crawling & Indexing • Relevance Ranking • PageRank • Early Quality Assessment 2010s - 2020s

If You’ve Been Doing Good SEO, This Isn’t New THE
AUTHENTIC SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR This isn’t an SEO session, but since 2013 and the Hummingbird update, search engines have been looking to rank based on”things” not “strings” This is when things such as topical authority, entities and semantics became more important, and Google didn’t need explicit keywords in order to serve the best results The principles of this are very much ingrained into AI search and how LLMs serve their results Passages with semantic relevance win over keyword stuffed URLs and obsession over word count

SEO?

Chapter 2 AI Search: Process & The RARER Methodology

Our RARER Methodology Has Never Been More Pertinent THE AUTHENTIC
SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR These principles still apply to AI search LLMs are looking to reward the same things as search engines and return the most credible results to users The process is very much the same even if the nuts and bolts behind this have become nuanced It’s not “all just SEO” but good organic search marketing hasn’t changed in principle Ignoring the noise and the acronyms is key to focussing on the right thing and driving growth from AI surfaces

Keyword Universe SEO & AI Search Audit Authority Building Content
Briefing Tech Implementation Crawl & Reaudit SEO Copywriting Impact Monitoring SEO Reporting Content Implementation AIR SEO (AI Ready) Content Marketing Research Phase Campaign Phase Monthly & Admin SEO Roadmap & Strategy Competitor Analysis UX & CRO

How The Methodology Has Evolved Research Adapt Refine Execute Report Keyword universe, search intent, fan-out queries from Google AI, LLM prompt audits (what do the chatbots already know about your brand) Technical fixes for crawlability and extractability. Semantic coverage, schema, structure, passage friendly formatting. Adapting keyword universe based on changes in demand and evolving fan out queries. Human led, E-E-A-T first content. Refined utilising brand voice, expertise and actual experience. The refinement layer is the bit that you cannot fake with AI. Depends on human intuition. Implement across surfaces. SEO and AI Search treated as one campaign. No duplicated effort. No siloed strategies. Track rankings, traffic, revenue. Track citations, prompts, AI mentions. Always tie back to commercial outcome.

Which Is Why We Created AIR SEO THE AUTHENTIC SEO
COMPANY | AI SEARCH MASTERCLASS | HIVEMCR It may sound gimmicky, but the prudent approach to AI Search Optimisation is AI-Ready SEO (AIR SEO) Not AI first, as this would create web pages that are agent friendly, not user friendly. Despite being pretty darn clever, agents are still nowhere near as clever as humans Remembering that all LLM results are probabilistic and not truly intelligent is vital But, by leveraging what we will discuss today, you can maintain and continue to improve your SEO performance and increase visibility in LLMs in a safe and futureproof way

Why Not AI First? THE AUTHENTIC SEO COMPANY | AI
SEARCH MASTERCLASS | HIVEMCR AI First • Optimising primarily for LLMs • Chasing a small section of total search demand (currently) • Potentially compromise UX and CRO • Selling clients a shiny new acronym they may not need AI Ready SEO • Solid SEO foundations, ready for AI retrieval • Capture traffic from traditional and AI surfaces • Human first content that converts • Baked into existing SEO campaigns, not a bolt on or upsell

The Google Stance THE AUTHENTIC SEO COMPANY | AI SEARCH
MASTERCLASS | HIVEMCR Google Says You Don't Need • LLMS.txt files • Special markup • Chunking content for AI • Rewriting content just for AI systems • Chasing inauthentic mentions • Over focusing on structured data What AIR SEO Focuses On • Useful content for real audiences • Crawlable, indexable foundations • Topical depth • Genuine authority • A brand worth surfacing • Fundamentals that drive revenue Disclaimer: Believing everything Google tells you can severely damage your visibility 😂

Updated Just Last Week - Approach With Caution Whilst in principle, there is nothing incorrect here, approach with caution. Things to consider: - Google isn’t the only LLM - They don’t use not to build links - We already know that creating content certain formats work - They’re already looking to stop people gaming the system - AI search has already opened the door for a lot of black hat techniques - LLMs.txt was in their documentation until recently - But the theme of concentrating on good SEO and good marketing practice, is sound

My Grandad, Frank Minor THE AUTHENTIC SEO COMPANY | AI
SEARCH MASTERCLASS | HIVEMCR Not Too Much Has Changed TECHNICAL FOUNDATIONS Crawlability, indexability, speed and structure INTRICATE RESEARCH Audience, market, intent, demand and user behaviour CONTENT STRATEGY Led by search data, accurate, compelling and written for humans PEOPLE FIRST MARKETING The user at the heart, agentic commerce may be coming but generally, people buy not machines BRAND & AUTHORITY BUILDING Authority, trust, recognition Expertise, reputation and advocacy ORGANIC SEARCH FUNDAMENTALS

Citation Factors Source: https://signal.zyppy.com/p/ai-citation-ranking-factors

Chapter 3 What Is The Retrieval Layer The Technical Stuff

The Retrieval Layer Needs Solid Legacy Search Foundations THE AUTHENTIC
SEO COMPANY | AI SEARCH MASTERCLASS | HIVEMCR • Strong traditional search foundations make sophisticated AI retrieval possible. • This is how agents and LLMs train themselves on what's worth surfacing. • Weak foundations, tend to result in weak AI search visibility • Strong foundations are already a fantastic start for AI search, due to ht • This is why many brands are performing well despite limited “GEO” AI Retrieval (Cited, Summarised, Surfaced) Content (Crawled & Indexed) Search Foundations (Authority, Structure, EEAT)

Retrieval Systems These modern retrieval systems use some classic ranking models and keyword matching, but with the flexibility of semantic search. BM25 (Best Matching 25): This is keyword based scoring, similar to how traditional search works. This algorithm considers keyword frequency and length of document, so assess how well it matches the query. This is likely to have quite a low weight, but we know that it is used. Vector Similarity Scoring: Measures how closely the semantic embeddings align with the query embedding, enabling conceptual matching. So means you can be sited even if your content doesnt match the keywords within the prompt exactly. RRF (Reciprocal Rank Fusion): This combines multiple ranking signals to get a final relevance score. This means that both keyword precision and semantics are at play, providing a hybrid approach. These, and likely many other algorithms that the LLMs don’t share all combine, meaning that both traditional relevance signals and semantic alignment are key. You need to consider both, which has been prudent for SEO performance for some time.

AI Systems: From Input to Understanding 1. Content Ingestion (Preparing It For Analysis) • Text extraction from HTML, PDF etc • Cleansing data • Establishing language and encoding • Content structure analysis (headings, lists etc) Clean HTML, markup, logical structure helps more accurate processing. Solid Technical SEO 2. Linguistic Analysis • Tokenisation • Grammar analysis • Entity recognition: people, places, organisations • Grammatical relationships between words Take this past what words are on the page and analyses how they relate to each other. Solid Technical SEO 3. Semantic Processing & Embeddings • Contextual embeddings • Chunk level embeddings • Topic modelling • Relationship extraction This determined how your content aligns with different queries and whether it should be retrieved for them. Solid Technical SEO 4. Storage & Indexing • Vector databases • Traditional indexes for keyword retrieval • Metadata for structure and authority information • Update mechanisms to keep this fresh All this needs to be stored, so it can be retrieved efficiently. Onus on LLM ,

AI Systems: From Input to Understanding 5. Query Processing & Retrieval • Query analysis • Utilising several retrieval methods • Ranking and fusion combines relevance signals • Context window management When queries arrive, the LLMs then match them against that stored data. This is what determines what content does well in LLMs and what doesn’t It is important to note that this isn’t definitive and each LLM will use different methods, but this is broadly how they operate and will continue to do based on all available documentation

Tokens & Context Windows The fundamental difference in the way that AI agents retrieve versus search engine bots, is tokens and context windows. Over optimisation over the years has led to something similar to this across the web: Heading Engaging Copy Filler Copy Engaging Copy Filler Copy But this approach will now effectively be “taxed” as LLMs need to operate using expensive tokens and therefore have a context window. Each chunk is a block of tokens they need to evaluate Useful, dense, well structured passages get extracted. Padding, or filler designed to hit a word count, doesn’t This doesn’t mean pages cannot be long, as some in-depth topics will need that. But fluff is now is more counter productive than ever

Tokens Understanding tokenisation is the first step, to understanding the difference between how your web pages are crawled and indexed and how your content is retrieved by LLMs. The first thing an LLM does is break your content down into tokens, and these should not be wasted. Remembering that all LLMs will be slightly different, have different token budgets and differing levels of resource overall. But the below is widely accepted for all the major surfaces 1 token = 0.75 words

We Can Break This Down Further English Text Standard Prose: 1,000 words equals about 1,300 to 1,500 tokens Short/Common Words: Common short words (e.g., "the", "and") are frequently assigned exactly 1 token Longer/Complex Words: Longer words (e.g., "unbelievable") are often split into 2 or 3 tokens Non-Standard Text Code: Code requires more tokens due to symbols, indentations, and syntax 1,000 words of code equals about 2,000 to 3,000 tokens JSON/XML: Structural characters add overhead, meaning 1,000 words of data equals about 3,000 to 4,000 tokens Other Languages: Languages like Chinese, Japanese, or languages with complex morphology will require more tokens per word than English.

Tokens This emphasises the need to be concise and avoid filler content when optimising for AI search • If LLMs frequently retrieve chunks that have no purpose they could lose trust in your brand • Don’t waste their tokens and be taxed for overly verbose content, that would have performed well for legacy SEO • Word count isn’t an SEO ranking factor, engaging and compelling content is Could AI search be the start of SEOs cleaning up all the mess they’ve made over the last 30 years? Think about the Wikipedia rule, where moderators insist in every passage of content needing a point and ideally, references and citations.

Context Windows This has created something called the context window, where you have limited time (and tokens) with which to make your point, or the likelihood is that your passage will not be considered for citation Things to consider aside from verbosity are: • Complex vocabulary and jargon (this pushes key info outside of the context window) • Sentence structure • Information hierarchy (tokens are processed sequentially - hence AI summaries etc)

Current Context Limits As each LLM is a different model, and a different business with different budgets and different levels of resource, they have different context limits. It isn’t necessarily that the bigger the context limit, the better the LLM. But it does tell you what they’re capable of (thinking about training data and grounded search)

Training Data vs Grounded Search THE AUTHENTIC SEO COMPANY |
AI SEARCH MASTERCLASS | HIVEMCR Feature AI Training AI Grounding What it is Building the model’s core brain, language skills, and general knowledge. Connecting the AI to specific, external facts (like your company's documents or live web results). How it works Ingesting billions of pages from the internet or datasets over a long, expensive industrial process. Providing the model with contextual evidence during your conversation, often via frameworks like Retrieval-Augmented Generation (RAG). Limitation Without it, the AI has no language or reasoning capability. Without it, the AI can suffer from "hallucinations" (making things up) or rely on outdated training information. Transparency It relies on statistical probabilities, making it difficult to trace where an exact fact came from. It connects abstract words to reality, often citing specific URLs, documents, or data sources.

Training Data vs Grounded Search THE AUTHENTIC SEO COMPANY |
AI SEARCH MASTERCLASS | HIVEMCR • This is where we can see if our data is being used for various LLM’s training data or grounded search • So effectively, if they’re retrieving your info to train, or pretty sure they can now answer questions based on it • And ideally, cite your domain

Chunking This seems to have become a dirty word in the world of AI search, as SEOs (and GEO bros) are already trying to manipulate it, but it isn’t. It is key Tokenisation ➡ Chunking Whilst we shouldn’t be going through all our old content and splitting pages into chunks (LLMs will do this anyway), using PeopleAlsoAsked data and Fan Out Queries, it needs consideration. All RAG systems (retrieval augmented generation), which is what powers most AI applications, work with chunks rather than passing entire documents into the system. Again a key difference between AI search and traditional search. A good way to think about this is whereby a human would read content continuously, and build understanding. AI works with these chunks, which are self contained and essentially isolated. So in some ways, not that intelligent at all and rather, probabilistic.

What is RAG (Retrieval Augmented Generation)?

Chunking Approach this as semantic completeness. Not every chunk needs to be designed for retrieval, but any that aren’t are unlikely to be surfaced. Headers and structure are just good SEO (again think Wikipedia), but especially helpful for LLMs as this can guide them in terms of how they chunk your content. But related information needs to be in or near the same chunk, if you explain a concept in one place and provide examples in another, the LLM is very unlikely to connect them and they could end up in different chunks. So content organisation does need a rethink in the post AI era, but it’s about finding the sweet spot between SEO and AI search given, until such a time that this is completely baked into Google’s algorithms.

Vector Embeddings This has been a technical SEO buzzword for a while, and rightly so as it can be super effective. Remembering that this means restructuring content purely for bots and agents, and not humans But this is the layer that allows AI systems to actually understand what your chunks of content actually mean. They are a mathematical representation of meaning, that allow the systems to understand relationships between concepts even if they don’t share the same terms - though the clustering of these concepts I.e Car, automobile, vehicle. This is the sort of technology that allowed Hummingbird to make the search results more semantic, and it very much how AI systems are powered in terms of understanding. This is when search went from strings, to things

How AI Systems Work So very broadly we have the following process to consider when creating content. Tokenisation > The Context Window > Chunking > Vector Embeddings > Citation This is all steeped in semantic search and finding content based on conceptual similarity rather than exact keyword matches. It’s how search has worked for over a decade (alongside a lot of other ranking factors), but the foundations of AI search. Things to remember: - Use Natural Language Variations (less keyword density, more synonyms) - Cover Concepts Thoroughly (several angles, no filler content) - Connect Related Concepts Explicitly (do this through your writing rather than technically) - Maintain Semantic Consistency (no mixed messages or ambiguity) All of this should be good for the user, so we can write human first content that is also super retrievable (if we’re clever)

Chapter 4 Utilising AI For Content Strategy

Optimising for Machines & Humans “How Do We Write for Machine Comprehension Without Losing Human Appeal?” We’ve already covered: • Structural clarity • Semantic richness …and these elements should ensure that content is user friendly rather than long form, over SEO’ed articles. And AI is also looking to reward • Factual accuracy • Verifiability • Advocacy All these things are more “human” than the traditional ranking factors.

Could AI Search Optimisation Help Clean Up SEO Mess? After two decades of word count obsession and keyword stuff, could AI search spur a clean up? Classic SEO Era • “Ultimate Guides” and filler content • Keyword density focus • SEO queries above the fold • Content written for bots, not people Post AI SEO Era • Concise, structured passages (think Wikipedia) • Direct answers and summaries above the fold • Clear entities and definitions • Depth, but only when depth is required

Content Architecture Finding the sweet spot between what is optimal for AI search versus traditional organic search is tricky, and can depend on the type of website Executive Summary Layers • Quick overviews, work well for AI synthesis • Help humans to understand content quickly • Can be harmful if you want users to interact below the fold with ads etc Detailed Explanation Layers • Covers all angles so i rewarded by LLMs • Can rank well for long tail SEO queries • Vital these do not contain waffly filler content Supporting Evidence Layers • Data, citations, examples, technical details • Reinforce authority • Specific information for retrieval systems Practical Application Layers • Connect concepts to real world usage • Case studies, examples and implementation • Aids traditional searches and retrieval systems

Other Content Architecture Techniques There are other ways you can technically and creatively look to optimise for both traditional search and retrieval systems. Modular Content Design • Self contained modules • Consistent formatting patterns • Cross reference systems to allow retrieval systems to understand the relationship between different pieces of content Authority Architecture • E-E-A-T demonstration such as author credentials, institutional affiliation and subject matter depth • Citation and reference architecture, to provide sourcing for claims and informational accuracy

Summary Ultimately, we do not need to write a whole new playbook for what good content strategy looks like, just because Google’s changed it’s algorithms and we have new LLM surfaces to optimise for. But we do need to pivot, and we’re used to that. A lot of these changes are positive. “If you’re an marketer who has enjoyed SEO, the future is bright, if you’re an SEO who has reluctantly embraced marketing, then the future could be tricky”

Chapter 5 Some Tips & Tools

AI Tooling Optimising your site to be cited in AI search is one thing, but we also need to consider how to safely use AI for our SEO campaigns. Some pitfalls. Pitfalls • Hallucinations (and poor prompting) • Content creation (incredibly hard to achieve excellence) • Technical SEO auditing (i’m still to see an accurate audit, and agents lack intuition) • Gaslighting (particularly OpenAI) Opportunities • Research at scale • Refining large data sets • Briefing documents (content and technical SEO) • Rapid trend analysis

AI Tooling Much like all the GEO bros and grifters, there has been a scramble to provide “AI visibility” and “AI citation tools. Which can be useful, but can encourage the wrong KPIs and are extortionately expensive.

AI Tooling A wiser investment could be to invest in tools that allow you to optimise, rather than focus on vanity metrics that may not even be driving growth. Citations are great, but we often don’t know the prompts, traffic and conversion is still king/queen. Let’s take a look.

What We’ve Seen From Our Own Campaigns Traffic from LLMs should still be a key KPI, over visibility and citation metrics as this will drive performance.

What We’ve Seen From Our Own Campaigns Correlates almost identically with SEO visibility as LLM visibility responds to optimisation. And with this, we can see what actually drives that visibility (rankings, keyword data). So using both sources holistically is prudent.

What Does This Tell Us? That the retrieval layer interacts with traditional rankings in practice Authority Amplification As LLMs are presenting information as facts rather than a list of links to choose from, they are most likely more concerned with authority. By this we don’t mean domain authority and backlinks, but more authoritative sources as the LLM does not want to hallucinate or give out false information. Comprehensive Content Long form, comprehensive content performs well due to covering topics from multiple angles. Provided this content is concise and not filler content then it will be rewarded by LLM for providing complete answers. Structured Content Boost Clear organisation, consistent formatting a logical hierarchy is rewarded in LLMs, even more so than in traditional search. This all supports the logic that the retrieval layer is an enhancement to the traditional rankings, rather than a replacement.

Chapter 6 Reporting & The Future

The Crocodile Effect Has Affected Many Sites We’ve probably all seen a bit of this, and there are a lot of factors driving this from • Increased Sponsored Results (likely more damaging to SEO performance than AI right now) • AI Overviews • AI Mode But, for brands that are optimising effectively, they may be seeing impressions increasing (across many different experiences), clicks dropping (zero click), but revenue not necessarily dropping at the same rate. GA tracking only tells us part of the story so tracking revenue versus performance across all organic channels is paramount

Reporting & Attribution Traditional Metrics Retrieval Performance Metrics Ranking Positions (can affect retrieval system candidate selection) AI Citation Frequency (how many of the AI tools measure success currently) Organic Traffic (can offer user behaviour data that influences authority assessment) Semantic Coverage Analysis Engagement metrics (content value) Cross Platform Retrieval Consistency (manual) Authority indicators (retrieval system trust assessment) Content Window Utilisation (content that delivers value in fewer tokens) Tools do not exist for all of these metrics yet but they will (or should)

Predictions for 2026 And Beyond Multi Modal Results Visual search, USG, mixed media SERPs, richer answer formats. Your content strategy needs to plan for more than text Hyper Personalisation Real-time adaptation to user context, history and Gmail level data signals. Let go of generic ranking position as your KPI. Relevance Over Volume Optimisation centred on real time relevance and learning, from both Google's AI layers and third-party tools. Search volume won’t become obsolete, but just part of the picture USG is huge, Google’s partnership with Reddit was not just for SEO reasons, remember that most LLMs have trained hugely on that data and continue to, despite it being riddled with misinformation But this is something we can control. Also remember that 90% of indexed Reddit threads are negative.

Final Summary E-E-A-T Principles 03 • Expertise • Experience • Advocacy Semantic Mapping 02 • Entities • Fan out queries • Passage level depth Foundations 01 • Technical SEO fundamentals are still key • Allow AI agents to crawl your site, ensure it is fast and well structured • Remember that most agents will not render JavaScript Redefine Your KPIs Leading With Revenue 06 • Revenue is king • Beware the crocodile effect • Get rid of vanity metrics, and focus on commercial outcomes AI Search Is a Layer of Organic Not a Channel 05 • Rankings are good insight, but not a KPI • Being cited can be as valuable • Report on on it all for a holistic view AI Tooling Are Your Assistants 04 • Research • Briefing • Analysis

AI Search Optimisation - HiveMCR Masterclass

AI Search Optimisation - HiveMCR Masterclass

More Decks by Whitworth SEO

Other Decks in Marketing & SEO

Featured

Transcript