Most checklists for search optimization were written for a world with ten blue links. They are still useful, but they miss what changes when an AI system reads your page and writes an answer from it. This checklist keeps the good habits of classic SEO and adds the checks that matter for AI answers. Use it before you publish a page, and again when you review old ones.
By the iSuggest.ai Team ยท Updated for 2026
It is grouped into five areas: access, content, structure, identity and measurement. Each check is short on purpose. If you want the reasoning behind a group, we link to a longer guide. If you prefer to have a tool run many of these checks for you, the free audit on iSuggest.ai does exactly that; see what an iSuggest.ai audit measures.
Why a checklist, and why now
The need is real. Industry compilations report that AI Overviews appear on about 48% of tracked queries and that zero-click searches make up roughly 68% of US Google queries (SEO Sherpa, Digital Applied). Meanwhile, traffic from AI search engines grew about 16x between 2024 and 2026. These figures differ by dataset, so treat them as direction, but the direction is consistent: more answers, fewer clicks, and more of your audience arriving through assistants. A repeatable checklist is the cheapest way to keep up. If you want the strategic picture, read GEO vs SEO: ranking is not enough.
Part 1: Access (can a machine reach the page?)
If a crawler cannot fetch your page, nothing else counts. These five checks come first.
- The page returns a 200 status. Redirect chains, soft 404s and server errors all reduce the chance of being read. Fetch the URL from outside your network and confirm.
- The page is not blocked by robots.txt. Read the file line by line, including wildcard rules. Common syntax mistakes block more than their authors intend.
- You have made a deliberate choice about AI crawlers. Know which user agents you allow and which you block, and why. Our post on the rise of AI crawlers lists the main ones.
- Your firewall or CDN does not challenge legitimate bots. A bot-protection page served in place of your content is what a crawler will read.
- The key content is in the raw HTML. View the page source, not the inspected DOM, and search for a sentence from your main copy. If it is missing, see what client-side rendering hides.
Part 2: Content (is the page worth quoting?)
Once a page can be read, the question is whether it is worth using. These checks are about substance and shape.
- The opening answers the page's main question. A reader, or a model, should get a correct answer from the first hundred words.
- Headings are real questions or clear topics. "How much does it cost?" beats "Overview".
- Each paragraph carries one idea. Short, self-contained paragraphs are easier to extract without losing meaning.
- Terms are defined where they first appear. A quoted passage loses its surroundings, so keep it understandable on its own.
- Claims are specific and sourced. Link to the original research and keep the figure's framing intact. Never round a correlation into a guarantee.
- The page adds something others do not. Original experience, data, examples or a clear point of view. A page that repeats what ten others say gives no reason to be chosen.
- Structure matches the content. Numbered steps for procedures, tables for comparisons, short definitions for terms. Do not force a list onto reasoning that needs prose.
For more on the writing side, see how to write content AI models want to quote.
Part 3: Structure (does the markup say what the page is?)
Structured data and clean markup remove ambiguity. These checks are cheap and effective.
- There is exactly one H1, and it matches the page topic. Heading levels should nest logically, without skipping.
- The title tag and meta description are unique and accurate. Avoid duplicates across pages and avoid doubling your brand suffix.
- The canonical URL is correct. It should point to the preferred version of the page itself. Canonical tags are easy to get subtly wrong.
- Structured data is present and valid. Use Organization on the homepage, Article on posts, and FAQPage or HowTo only where the visible content matches. See structured data for AI search engines and the FAQ schema guide.
- Images have meaningful alt text. Describe what the image shows, not keywords you hope to rank for.
- Links use descriptive anchor text. "Read our pricing guide" tells a crawler more than "click here".
Part 4: Identity (does the web agree on who you are?)
An AI system has to be confident about the entity behind a page. These checks build that confidence.
- Your name and description are consistent everywhere. Site, profiles, directories and press mentions should describe you the same way. See entity clarity.
- There is a clear About page and real authorship. Name the people or the organization responsible for the content.
- Official profiles are linked, including through sameAs markup. This connects your site to the rest of your presence.
- You are building credible mentions elsewhere. One analysis reports brand mentions correlate with AI visibility about three times more strongly than backlinks (0.664 versus 0.218, per Media Officers). That is a correlation, not a guarantee, but it supports investing in coverage; earned media beats press releases by a wide margin in the citation data.
Part 5: Measurement (will you know if it worked?)
- You have a baseline. Run an audit now and record the score and the top issues. Without a starting point, you cannot see progress.
- You track the right numbers. Watch organic impressions and clicks in Search Console, referral sessions from AI assistants in your analytics, and branded search over time.
- You re-check on a schedule. Our recommendation is in how often to re-audit your website.
The most common failures we see
Across pages of every kind, the same handful of problems turn up again and again. Knowing them in advance helps you prioritize.
Blocks nobody remembers adding
A plugin, a security rule or a developer's old fix adds a Disallow line or a firewall rule, and years later it is still there. The symptom is a page that is perfectly good and quietly unreachable. Re-read your robots.txt twice a year and ask your host what bot protection is switched on.
Content that exists only after JavaScript runs
Modern frameworks make this easy to do by accident. Single-page applications often ship an empty shell and fill it in the browser. A human sees a full page; a fetch that does not run scripts sees almost nothing. Checking the page source takes ten seconds and settles it.
Openings that never answer
Many pages spend the first screen on a slogan, a hero image and a button. That is fine for a landing page, but a page that is meant to be quoted needs a plain statement of what it is and what it says near the top. If your best sentence is on the third scroll, move it up.
Markup that contradicts the page
FAQ markup for questions that are not on the page, review markup with no visible reviews, and Organization data that disagrees with the footer all create doubt. Structured data should be a faithful label, never an embellishment.
Inconsistent names
If you are "Acme Inc." on your site, "Acme Software" on your profile and "ACME" in a directory, you make a model's job harder. Pick one name and one description and use them everywhere.
How to prioritize when time is short
Not every check is equal. If you can only do a few things, rank them by how much they unblock. Access first, because a blocked page cannot benefit from anything else. Then the opening answer and the heading structure, which affect how much of the page can be reused. Then structured data and identity signals, which affect how confidently a system can attribute the content. Measurement comes last in this list but should start on day one, because a baseline recorded now is the only way to show progress later.
A sensible monthly rhythm is to audit your five most important pages, fix the three highest-impact items, and note the change. In a quarter you will have fixed fifteen real problems and have a score history to show for it. That compounding habit beats a single heroic cleanup that is never repeated.
Who should own this checklist
On small teams, it tends to fall between roles. The developer owns robots.txt and rendering, the writer owns openings and headings, the marketer owns mentions and profiles, and nobody owns the whole. Give one person the job of running the checklist each month and passing items to the right owner. For agencies running this across client sites, the same logic applies at larger scale, and a standard audit report makes handoffs far easier.
It also helps to add the checklist to your publishing workflow so new pages meet it from day one. Retrofitting is always more work than getting it right at launch. A short pre-publish step, running through parts one to three, catches most problems before the page ever goes live. Our guide to content refresh at scale explains how to apply the same checks to a back catalogue without rewriting everything by hand.
How to use this checklist without drowning
Twenty-five checks is a lot if you try to do them all at once on every page. A simpler approach is to apply the full list to your five most important pages first: the homepage, your main product or service page, your pricing page, and your two most visited articles. Fix the access issues before anything else, because they decide whether the rest is even visible. Then work through content, structure, identity and measurement in order.
For most sites, the quick wins cluster in the same places. Robots rules inherited from an old setup. Content injected by scripts. Missing or inconsistent structured data. An opening paragraph that takes too long to say anything. None of these needs a redesign, and each can be fixed in an afternoon.
What a checklist cannot do
A checklist improves your odds; it does not decide the outcome. Whether an assistant cites you depends on factors nobody outside the model providers can see, and those factors shift over time. Be skeptical of anyone who promises guaranteed placement. What you can promise yourself is a site that is reachable, readable, well described and kept up to date, which is good for readers regardless of what any model does.
Turn the list into a score
Reading a checklist is easy; applying it page by page is slow. If you would rather see where you stand in a minute, see how an iSuggest.ai audit works and run it on your own URL. The audit is free and not metered, so you can check as many pages as you like. When a page is ready, you can create an account at your account page and publish a snapshot to the ChatGPT directory or the others; publishing spends credits and improves findability without promising a citation, as explained in our credit system guide.