What Makes a Website Easy for AI Tools to Understand: The Complete Guide to AI-Readable Websites

Your website has a new visitor. It doesn't see your hero image, your brand colors, or your carefully crafted hover animations. It reads your raw HTML, line by line, looking for structure, meaning, and extractable facts. This visitor is an AI agent — a large language model like ChatGPT, Gemini, Claude, or Perplexity — and it is rapidly becoming the most important gatekeeper between your content and your audience.

According to McKinsey, 50% of consumers now use AI-powered search. Bain reports that 60% of traditional search queries end without a single click, as users get their answers directly from AI-generated summaries. Google processes 480 trillion tokens through its AI models every month — a 50x increase over the previous year. The implications are stark: if your website isn't readable by AI tools, you are invisible to a growing share of your potential audience.

But what actually makes a website easy for AI tools to understand? Is it a secret algorithm, a proprietary file format, or a new kind of optimization entirely separate from traditional SEO? The answer is both simpler and more demanding than most people expect. AI readability is not a new discipline — it is the convergence of accessibility, semantic HTML, structured data, and clear content architecture. The same principles that make your site work for screen readers make it work for AI agents. The same discipline that earned you featured snippets in Google now earns you citations in AI Overviews.

This guide breaks down exactly what AI tools look for when they parse your website, the ten pillars of an AI-readable site, and a practical implementation roadmap you can follow starting today.

How AI Tools Actually Read Websites

Before optimizing for AI, you need to understand how AI systems parse web content. They don't browse the way humans do — and they don't crawl the way traditional search engine bots do, either.

Retrieval-Augmented Generation: The AI Reading Pipeline

When an AI assistant like ChatGPT or Perplexity answers a user's question, it typically uses a process called Retrieval-Augmented Generation (RAG). The system does not search the entire web in real time. Instead, it:

1. Receives a query and determines whether it needs external information.
2. Dispatches a retriever — a specialized tool that searches indexed web content or performs live lookups — to find relevant pages.
3. Extracts and chunks the most relevant passages from those pages into digestible segments.
4. Feeds those chunks into the language model as context, alongside the original query.
5. Generates a response that synthesizes the retrieved information, often with citations.

This pipeline has profound implications for how you should structure your website. The AI never sees your homepage as a whole. It sees fragments — paragraphs, headings, lists — pulled out of context and reassembled. If those fragments don't make sense independently, the AI will either misrepresent your content or skip your site entirely.

Chunking: Why Structure Matters More Than Ever

Chunking is the process of breaking long documents into smaller, overlapping segments that fit within a model's context window. A typical chunk might be 300 to 1,000 tokens. The AI's retriever decides which chunks are relevant to the query and passes only those to the model.

If a single idea is spread across three paragraphs, separated by a sidebar or an image, it may get split across different chunks — and the AI will lose the thread. If your headings are vague ("Learn More," "Details," "Services"), the retriever has no semantic signal to anchor on. Self-contained, clearly labeled content blocks are not just a readability preference; they are a structural necessity for AI comprehension.

Tokenization and Semantic Meaning

LLMs process text by breaking it into tokens — subword units that may be whole words, parts of words, or punctuation. The model then analyzes relationships between tokens to understand meaning. This is why consistent terminology matters. If you call your product a "platform" on one page, a "solution" on another, and a "tool" on a third, the AI's entity model fragments. It cannot build a stable understanding of what you offer.

The 10 Pillars of an AI-Readable Website

1. Semantic HTML: The Foundation of Machine Readability

Semantic HTML is the single most important factor in AI readability. It is the difference between handing the AI a labeled map and dropping it into a maze.

When you use generic <div> and <span> elements for everything — a practice known as "div soup" — you give the AI no structural cues. Every element looks the same. The AI cannot tell where the navigation ends and the main content begins, or whether a block of text is a sidebar, a footer, or the article body.

Semantic HTML5 elements solve this by explicitly labeling the role of each section:

• <header> — the page or section header, typically containing the logo, navigation, and introductory content
• <nav> — navigation links, clearly distinguished from body content
• <main> — the primary content of the page, exactly one per page
• <article> — a self-contained composition, such as a blog post or product description
• <section> — a thematic grouping of content, typically with its own heading
• <aside> — content tangentially related to the main content, such as sidebars
• <footer> — the page or section footer

When an AI retriever encounters these elements, it can immediately build a structural map of the page. It knows which content is primary and which is supplementary. It can prioritize <main> over <aside> when extracting answers. This is not a theoretical advantage — it directly affects which passages get chunked and fed into the model.

What to do: Audit your site's HTML. Replace generic containers with semantic elements wherever possible. Ensure every page has exactly one <main> element and that <article> or <section> wraps each distinct content block.

2. Heading Hierarchy: The AI's Table of Contents

AI models use heading tags (<h1> through <h6>) to build a mental outline of your page. A clear, logical heading hierarchy is one of the strongest signals you can send about content organization.

The rules are straightforward but frequently violated:

• One <h1> per page. This is the page's primary title and should clearly state what the page is about. It should include your primary keyword.
• <h2> elements define major sections. Think of them as chapter titles. Each should be descriptive and specific — not "Services" but "Cloud Migration Services for Enterprise."
• <h3> elements define subsections. They should nest logically under their parent <h2>.
• Never skip levels. Don't jump from <h1> to <h3>. The AI uses the hierarchy to understand parent-child relationships between topics.

When headings are just styled text — a <span> with a CSS class that makes it look big and bold — the AI sees no structure at all. It cannot differentiate a section title from body copy. This is one of the most common and costly mistakes in AI website optimization.

What to do: Run a heading audit using a tool like the W3C Validator or a browser extension that exposes the heading outline. Fix any skipped levels, missing headings, or non-semantic heading substitutes. Ensure every heading describes the content that follows.

3. Structured Data and Schema Markup: Speaking the AI's Native Language

If semantic HTML gives the AI a map of your page, structured data gives it a dictionary. Schema.org markup — implemented as JSON-LD, Microdata, or RDFa — explicitly labels the entities and relationships on your page. It tells the AI: "This is a product. Its price is $49.99. It has 4.3 stars based on 237 reviews. It is in stock."

A 2025 analysis of 1,500 websites found that 70% had zero schema markup. Of the 30% that did, 94% were using only basic organization schema. This represents an enormous competitive gap.

The schema types most relevant to AI understanding include:

• Organization — your brand name, logo, social profiles, and contact information. This is the minimum viable schema for any site.
• Article / BlogPosting — headline, author, date published, date modified, and publisher. Essential for any content-driven site.
• FAQ — question-and-answer pairs. This is arguably the highest-leverage schema for AI search, as it provides ready-made, extractable Q&A blocks that AI assistants can present verbatim.
• Product — name, description, price, availability, SKU, and aggregate rating. Critical for e-commerce.
• LocalBusiness — address, hours, phone, geo-coordinates, and accepted payment methods. Essential for local visibility.
• BreadcrumbList — the page's position in the site hierarchy. Helps AI understand site structure.
• HowTo — step-by-step instructions with materials, tools, and estimated time. Highly extractable for AI summaries.
• Review — item reviewed, rating, author, and review body. Provides social proof that AI can cite.

What to do: Implement JSON-LD structured data on every page. Start with Organization and Article schema. Add FAQ schema wherever you have question-and-answer content. Use Google's Rich Results Test to validate your markup. Never stuff schema with content that doesn't appear on the visible page — this is a violation of Google's guidelines and erodes trust.

4. Clear, Chunk-Friendly Content Architecture

Content structure is where SEO and AI readability converge most directly. The same formatting that earned you featured snippets in traditional search — the "inverted pyramid" style, clear subheadings, bulleted lists — now determines whether an AI model will cite you.

AI models prefer content that is:

• Front-loaded with answers. Put the most important information in the first 100 to 200 words. AI retrievers prioritize the opening paragraphs when extracting answer snippets. If your introduction is a meandering anecdote, the AI will likely skip it.
• Organized into self-contained paragraphs. Each paragraph should express one complete idea. If a paragraph depends on the one before it to make sense, it may fail when chunked independently.
• Formatted with lists and tables. AI models extract structured formats with high fidelity. A comparison table or a numbered list of steps is far more likely to appear in an AI-generated answer than a dense paragraph of prose.
• Dense with specifics. Concrete numbers, dates, percentages, and named entities are highly citable. "We reduced customer churn by 40% in 2025" is valuable to an AI. "We dramatically improved customer satisfaction" is not.
• Free of jargon and marketing fluff. AI models struggle with vague, hyperbolic language. "Leveraging synergistic paradigms to actualize next-generation solutions" conveys nothing retrievable. Clear, direct language wins.

What to do: Review your top 20 pages through the lens of chunking. Can each paragraph stand alone? Does the first paragraph answer the page's primary question? Are there lists and tables where they would add clarity? Rewrite any content that fails these tests.

5. Crawlability and Technical Health

An AI cannot understand content it cannot reach. Crawlability — the set of technical factors that determine whether automated systems can access your pages — is the prerequisite for everything else in this guide.

The fundamentals are well-established but frequently neglected:

• Successful HTTP responses. Every important page should return a 200 status code. Redirect chains, 404s, and 500 errors waste crawl budget and prevent content from being indexed.
• A clean, up-to-date XML sitemap. Submit it to Google Search Console and reference it in your robots.txt file. Include only canonical, indexable URLs.
• A properly configured robots.txt. Block pages you genuinely don't want crawled (admin panels, staging environments, internal search results). Never accidentally block your main content. Test with Google's robots.txt Tester.
• Canonical tags. Use <link rel="canonical"> to consolidate duplicate or similar pages. Without canonicalization, AI retrievers may encounter multiple versions of the same content and dilute their understanding.
• No orphan pages. Every important page should be reachable through internal links. Pages with no inbound links are invisible to crawlers and AI retrievers alike.

What to do: Run a full crawl of your site using a tool like Screaming Frog or Sitebulb. Fix every 4xx and 5xx error, eliminate redirect chains, and ensure every canonical tag points to the correct URL. Review your robots.txt for accidental blocks.

6. Internal Linking and Information Architecture

Internal links are relationship statements. They tell AI systems which pages are related, which topics are central, and how your site's knowledge is organized.

A strong internal linking strategy does three things for AI readability:

• Establishes topic clusters. When you link from a pillar page to multiple supporting pages — and back again — you create a semantic web that helps AI understand the breadth and depth of your coverage on a topic.
• Provides context through anchor text. Descriptive anchor text ("our guide to schema markup implementation") tells the AI what the linked page is about before it even fetches the content. Generic anchors ("click here," "learn more") waste this signal.
• Distributes crawl depth. Shallow site architecture — where every page is reachable within three to four clicks from the homepage — ensures AI retrievers can discover all your content efficiently.

Flat architecture is particularly important for AI retrieval. When an AI dispatches a retriever, it typically has a limited budget of pages to fetch. If your most valuable content is buried six levels deep in a complex hierarchy, it may never be reached.

What to do: Map your site's content into topic clusters. Ensure every page links to its parent topic and to related subtopics. Use descriptive, keyword-rich anchor text. Flatten your architecture so that all important pages are within three clicks of the homepage.

7. Entity Consistency: Say It the Same Way, Every Time

AI systems build entity models — internal representations of the people, organizations, products, and concepts you discuss. When your site uses inconsistent terminology for the same entity, the AI's model fragments. It cannot confidently link references together.

For example, if your company is called "Lookelo" on your homepage, "Lookelo Inc." in your schema, "Lookelo Digital" in your blog, and "The Lookelo Team" in your about page, the AI sees four separate entities — or worse, it sees confusion and treats your brand as unreliable.

Entity consistency applies to:

• Brand names. Use the exact same name everywhere, including in schema markup, social profiles, and third-party directories.
• Product names. Never alternate between "Pro Plan," "Professional Tier," and "Premium Subscription" for the same offering.
• Locations and addresses. Format addresses consistently. Don't abbreviate "Street" on one page and spell it out on another.
• People. Author names, executive bios, and team pages should use consistent full names.

This principle extends beyond your own site. AI systems cross-reference information across the web. If your Google Business Profile, LinkedIn page, and website all use slightly different versions of your business name, the AI's confidence in your entity drops. Yext and similar platforms exist specifically to solve this problem by synchronizing entity data across the web.

What to do: Create a brand style guide that specifies exact entity names. Audit your site for inconsistencies. Update your schema markup, social profiles, and third-party listings to match. Use the same naming conventions everywhere.

8. Server-Side Rendering and JavaScript Considerations

If your website relies entirely on client-side JavaScript to render content, an AI crawler may see an empty page. This is one of the most technical — and most damaging — AI readability failures.

Traditional search engines like Google have invested heavily in rendering JavaScript. Their crawlers can execute client-side code and index dynamically loaded content — though even Google acknowledges this process is slower and less reliable than crawling static HTML. AI retrievers, however, are generally less sophisticated. Many of them fetch the raw HTML response and parse it without executing JavaScript at all.

If your content is loaded via API calls after the page renders — common in single-page applications built with React, Vue, or Angular — that content may be invisible to AI tools.

The solution is server-side rendering (SSR) or static site generation (SSG):

• SSR renders the page on the server and sends complete HTML to the client. The initial response contains all the content, readable by any crawler or retriever.
• SSG pre-renders pages at build time, producing static HTML files. This is the most AI-friendly approach, as the content is immediately available with no computation required.
• Hydration — where a framework takes over after the initial HTML load — is fine as long as the content is present in the initial HTML payload.

What to do: If you use a JavaScript framework, verify that your content appears in the raw HTML source (View Source, not Inspect Element). If it doesn't, implement SSR via Next.js, Nuxt, SvelteKit, or a similar framework. For content sites, consider SSG. Test by fetching your pages with a tool like curl or by viewing them in a text-only browser.

9. Accessibility as AI-Readability: The Screen Reader Parallel

There is a useful rule of thumb: if your website works well for a human using a screen reader, it almost certainly works well for an AI crawler. The overlap between accessibility and AI readability is not coincidental — both depend on the same underlying signals.

The key accessibility practices that directly improve AI understanding include:

• Descriptive alt text for images. AI cannot "see" images, but it can read alt text. Every meaningful image should have alt text that describes its content and function. Decorative images should use empty alt attributes (alt="") to signal that they can be ignored.
• Proper label associations for form elements. AI agents that interact with forms (for booking, purchasing, or subscribing) rely on proper <label> associations to understand what each field expects.
• Descriptive link text. "Download the 2025 AI readability report" tells the AI exactly what to expect. "Click here" tells it nothing.
• ARIA landmarks. When semantic HTML elements aren't sufficient, ARIA roles (role="banner", role="navigation", role="main") provide additional structural cues.
• Sufficient color contrast and readable font sizes. While these primarily benefit human users, they correlate with overall content quality — a signal AI systems increasingly consider.

What to do: Run an accessibility audit using Lighthouse, Axe, or WAVE. Fix all critical and serious issues. Treat accessibility compliance not as a legal checkbox but as a core component of your AI readiness strategy.

10. llms.txt and Emerging AI-Specific Standards

The newest entry in the AI readability toolkit is llms.txt — a proposed standard for a plain-text file placed at the root of a website that provides a curated overview of the site's most important content for AI agents.

Unlike robots.txt, which controls crawler access, llms.txt is a guide. It is written in Markdown and typically includes:

• A brief description of the site and its purpose
• Links to key pages, with short descriptions of each
• Pointers to documentation, APIs, or structured data endpoints
• Any information that would help an AI agent understand and navigate the site efficiently

A simple llms.txt might look like this:

# Lookelo
> SEO and AI search visibility agency helping businesses get found on Google and recommended by AI assistants.

## Key Pages
- [Services](/services): SEO, AI search visibility, and reporting offerings
- [AI Search Visibility](/ai-search-visibility): How we get brands cited by ChatGPT, Gemini, and Perplexity
- [Blog](/blog): AI search, GEO, and SEO insights
- [About](/about): Who we are and how we work

The standard is still emerging. Google has acknowledged that llms.txt can help agents understand site structure but does not currently require it for AI features. There is limited public evidence that simply adding the file increases citations. It is best viewed as a helpful supplement to the other nine pillars, not a replacement for any of them.

A related development is the Model Context Protocol (MCP), a more ambitious framework that allows AI agents to interact with structured data endpoints directly, rather than parsing HTML. While MCP is still in early adoption, it points to a future where AI-specific data access layers become standard infrastructure.

What to do: Add an llms.txt file to your site's root directory. Keep it concise — link to your most important pages, not every page. Update it when your site structure changes. Consider it a low-effort, moderate-reward optimization.

Common Mistakes in AI Website Optimization

Even well-intentioned teams make predictable errors when optimizing for AI readability. Here are the most frequent and damaging:

Mistake	Why It Hurts	How to Fix It
Client-side rendering without SSR	AI retrievers see empty pages; content is invisible	Implement SSR or SSG; verify content in raw HTML source
Vague headings ("Services," "Learn More")	No semantic signal for the AI's structural map	Use descriptive, keyword-rich headings that summarize the content
Missing or incorrect schema markup	AI has no explicit entity labels to work with	Implement JSON-LD for Organization, Article, FAQ, and Product as applicable
Content hidden in images or PDFs	AI cannot extract text from non-text formats	Publish all key information as HTML text; use images for decoration, not content
Inconsistent entity names	AI's entity model fragments; confidence drops	Standardize brand, product, and person names across all pages and platforms
Thin or duplicate content	AI has no unique value to extract; content is ignored	Consolidate thin pages; add original insights, data, and examples
Blocking important pages in robots.txt	Content is never crawled and never cited	Audit robots.txt; test with Google's robots.txt Tester
Over-reliance on llms.txt without fundamentals	The file is a map to a site the AI still cannot parse	Fix semantic HTML, schema, and content structure first; add llms.txt as a supplement

Traditional SEO vs. AI Readability vs. GEO: Understanding the Overlap

A common question is how AI readability relates to traditional SEO and Generative Engine Optimization (GEO). They are overlapping but distinct disciplines, and understanding the differences prevents wasted effort.

Factor	Traditional SEO	AI Readability	GEO (Generative Engine Optimization)
Primary goal	Rank in search results	Be correctly parsed and understood by AI	Be cited in AI-generated answers
Key techniques	Keywords, backlinks, page speed, content quality	Semantic HTML, schema, heading hierarchy, chunk-friendly content	All AI readability techniques plus entity authority, citation signals, and brand presence
Success metric	Organic rankings and click-through rate	Successful retrieval and accurate representation	Citations and brand mentions in AI responses
Relationship	Foundation for both	Prerequisite for GEO	The outcome of strong SEO + AI readability

The progression is linear: you cannot achieve GEO without AI readability, and you cannot achieve AI readability without strong technical SEO fundamentals. Building an AI-friendly website is not about abandoning SEO — it's about extending it to meet the needs of a new class of information consumer.

A Practical Implementation Roadmap

If you're optimizing an existing website for AI readability, here is a priority-ordered implementation plan:

Week 1–2: Technical Foundation

• Run a full crawl of your site and fix all 4xx/5xx errors
• Audit your robots.txt and XML sitemap
• Verify that all important content is in the raw HTML source (not loaded via JavaScript)
• Implement Organization and Article schema on all key pages

Week 3–4: Structural Improvements

• Audit and fix your heading hierarchy across all pages
• Replace generic <div> containers with semantic HTML5 elements
• Add descriptive alt text to all meaningful images
• Implement FAQ schema on pages with question-and-answer content

Week 5–6: Content Optimization

• Rewrite introductions to front-load key information
• Break long paragraphs into self-contained units
• Add lists and tables where they improve clarity
• Standardize entity names across all pages

Ongoing

• Add an llms.txt file and keep it updated
• Monitor AI citation tools (such as AmICited) to track whether your brand is being referenced
• Run regular accessibility audits
• Stay current with schema.org vocabulary additions and emerging AI standards

FAQ

How does semantic HTML help AI understand a website?

Semantic HTML uses purpose-built elements like <article>, <nav>, <main>, and <section> instead of generic <div> tags. These elements explicitly label the role of each content block, allowing AI retrievers to build a structural map of the page. The AI can distinguish primary content from navigation, sidebars, and footers, and prioritize the most relevant sections when extracting answers. Without semantic HTML, every element looks identical to the AI, and it cannot determine which content is important.

What is the difference between AI readability and traditional SEO?

Traditional SEO focuses on ranking in search engine results pages through keyword optimization, backlinks, and technical factors. AI readability focuses on making your content correctly parsed, chunked, and understood by language models. AI readability is a prerequisite for appearing in AI-generated answers — it's the layer beneath GEO. The two disciplines overlap significantly (both depend on crawlability, structured data, and content quality), but AI readability adds specific requirements around semantic HTML, self-contained content blocks, and entity consistency.

What schema markup types are most important for AI understanding?

Organization schema is the minimum viable markup for any site — it tells the AI who you are. Article and BlogPosting schema are essential for content-driven sites. FAQ schema is the highest-leverage type for AI search, as it provides ready-made Q&A pairs that AI assistants can extract and present verbatim. For e-commerce, Product schema with price, availability, and aggregate rating is critical. For local businesses, LocalBusiness schema ensures your address, hours, and contact information are machine-readable.

How do I implement an llms.txt file for AI agents?

Create a plain-text file named llms.txt in your site's root directory (e.g., yoursite.com/llms.txt). Write it in Markdown format. Include a brief description of your site, then list your most important pages with short descriptions. Keep it concise — the goal is to guide AI agents to your key content, not to catalog every page. You can also create a more detailed version at llms-full.txt that includes full Markdown versions of your key pages. Update the file whenever your site structure changes significantly.

Does page speed affect how AI tools understand a website?

Page speed affects AI readability indirectly. AI retrievers typically have timeout limits and crawl budgets. Slow-loading pages may not be fully fetched before the retriever moves on, resulting in incomplete content being indexed. Additionally, speed is a signal of overall site quality, which correlates with the authority signals AI systems increasingly consider. Fast, well-optimized pages are more likely to be fully retrieved and accurately parsed than slow, bloated ones.

What are the most common mistakes in AI website optimization?

The most common mistakes are building sites entirely with client-side JavaScript without server-side rendering, using vague headings instead of descriptive ones, hiding key information in images or PDFs, publishing thin or duplicate content, maintaining inconsistent entity names across pages, and accidentally blocking important pages from crawlers in robots.txt. Many teams also over-prioritize llms.txt while neglecting the foundational work of semantic HTML, schema markup, and content structure.

How does Google's AI understand websites differently from ChatGPT?

Google's AI systems (including AI Overviews and AI Mode) benefit from Google's existing search index, which has already crawled, rendered, and ranked your pages. They have access to structured data, PageRank signals, and the broader knowledge graph. Standalone AI assistants like ChatGPT use external retrievers that fetch pages on demand, often with less sophisticated rendering capabilities and no prior knowledge of your site. Both depend on the same fundamentals — semantic HTML, clear structure, and extractable content — but Google's systems are generally more forgiving of technical imperfections because they see your site through a richer indexing pipeline.

How do I optimize an e-commerce site for AI search?

E-commerce AI optimization starts with Product schema markup including name, price, availability, SKU, and aggregate rating on every product page. Implement BreadcrumbList schema to show category hierarchy. Use clear, descriptive product titles and specifications in structured formats (tables, lists). Ensure product descriptions are self-contained and include concrete details. Add FAQ schema for common customer questions about shipping, returns, and product compatibility. Maintain consistent product names across your site, Google Merchant Center, and third-party marketplaces. Ensure your product pages are rendered server-side so AI retrievers can access the full content.

What is the relationship between accessibility and AI readability?

Accessibility and AI readability share the same technical foundation. Descriptive alt text, proper heading hierarchies, semantic HTML, descriptive link text, and ARIA landmarks — all core accessibility requirements — are also the primary signals AI systems use to understand page structure and content. A site that is fully WCAG-compliant is typically 80% of the way to being AI-optimized. The alignment is so strong that a practical test for AI readiness is to navigate your site with a screen reader: if a human can understand the page through audio alone, an AI can almost certainly parse it correctly.

Conclusion

The websites that AI tools understand best aren't the ones chasing the newest file format or the most exotic schema type. They are the ones that get the fundamentals right: clean, semantic HTML; a logical heading hierarchy; self-contained, answer-rich content; consistent entity naming; and robust technical infrastructure.

This is ultimately good news. It means AI readability is not a separate discipline requiring a separate budget. It is the natural extension of web standards, accessibility, and content strategy that many teams are already practicing. The shift is in perspective: you are no longer optimizing solely for a search engine's ranking algorithm. You are optimizing for a machine reader that will extract, chunk, summarize, and cite your content — and that reader rewards clarity, structure, and specificity above all else.

Start with the technical foundation. Move to structure. Then optimize content. Add llms.txt as a finishing touch, not a starting point. Monitor your brand's presence in AI-generated answers using citation tracking tools. And remember: if a screen reader can understand your website, an AI agent probably can too. The two audiences, human and machine, converge on the same timeless principle — make your content clear, make it accessible, and make it true.