AI overviews show up at the top of Google, Perplexity, ChatGPT, and Bing. They summarize a topic, cite their sources, and push the regular links down the page. If your site is not included in the summary, you lose that click to a competitor who is.
Most advice on how to appear in these overviews boils down to "create great content" and "add schema." That advice is not wrong, but it is incomplete. Schema alone does not move citation rates, and "great content" does not help if a language model cannot extract your answer from it.
The IAB published a framework called "Measuring Visibility in the AI Era" on August 3, 2026. It breaks AI visibility into four dimensions: Presence, Prominence, Portrayal, and Persuasion. This framework maps directly to the signals that decide whether your site gets cited. Here is how each one works and what to fix.
How AI overviews actually work
Every major AI search engine runs the same basic pipeline. When someone asks a question, the engine runs a web search, pulls a set of candidate pages into working memory, reads each page, and synthesizes an answer with citations.
Google AI Overviews, Perplexity, and ChatGPT Search all follow this retrieval-augmented generation pattern. Google uses its own search index. Perplexity uses a mix of its own crawl and third-party indexes. ChatGPT uses Bing's index plus its own browsing. The indexes differ, but the pipeline is the same: find pages, read pages, write an answer, cite sources.
Citation is the last step in that chain. If your page fails at any earlier step, it never reaches citation. Most sites that do not get cited fail at retrieval or extraction, not at the final synthesis step.
The four signals that determine whether you get cited
The IAB framework defines four dimensions of AI visibility. Each one corresponds to a specific gate your content has to pass.
Presence: can the AI find your page?
Presence is the foundation. If an AI crawler cannot reach your page, nothing else matters. Your site is invisible to that engine.
Each AI search engine has its own crawler. OpenAI uses GPTBot. Perplexity uses PerplexityBot. Anthropic uses ClaudeBot. Google uses Googlebot for crawling and a token called Google-Extended to control whether your content feeds AI Overviews and Gemini.
Your robots.txt file controls access to all of them. If it blocks a crawler, that engine cannot read your site. Industry research suggests that many sites block AI crawlers accidentally, through default WordPress security settings, inherited templates, or Cloudflare bot-fight modes that sweep up AI crawlers alongside actual threats.
The fix is straightforward. Check your live robots.txt file at yourdomain.com/robots.txt. Read every User-agent and Disallow line. If you see a block for GPTBot, PerplexityBot, ClaudeBot, or Google-Extended, decide whether you want that block. For most businesses that depend on search visibility, the answer is no.
You can test crawler access directly with curl:
curl -A "GPTBot" -I https://yourdomain.com/A 200 response means the crawler can reach you. A 403 means something is blocking it.
Prominence: does the AI pick your page from the candidate pool?
Once your page is crawlable, it has to rank well enough to enter the candidate pool for a given query. AI search engines pull from the top of their indexes. Pages that rank in the top organic results get retrieved more often. Pages buried on page three rarely make it in.
This means traditional ranking still matters. Backlinks, topical authority, and content relevance determine whether your page gets retrieved. If you are not ranking, the AI engine never pulls your page in the first place.
But ranking is necessary, not sufficient. A page can rank in the top five and still not get cited. That happens at the next gate.
Portrayal: can the AI extract a clean answer from your page?
Portrayal is about accuracy and context. When the AI model reads your page, can it find a specific, extractable answer to the question?
This is where most sites lose. Your page ranks well. The model retrieves it. Then it reads your page, struggles to find a direct answer, and moves on to a cleaner page.
The pages that win here share specific characteristics. They state the answer in the first sentence under a heading that mirrors the question. They include specific data points, named sources, and concrete examples. They use clean semantic HTML with one H1 per page and a logical heading hierarchy. Short paragraphs. Clear structure. No important text trapped in images or loaded by JavaScript after the initial page response.
The model reads pages the way an impatient researcher does. It scans for the most direct, citable answer. If your page makes the model work to find the answer, and a competing page states it immediately under a matching heading, the competing page wins.
Persuasion: does your citation drive action?
Persuasion is the outcome dimension. When the AI cites your page, does the user click through? This depends on how your brand is portrayed in the answer and whether the citation links to a page that delivers on the promise.
Perplexity puts clickable source links inline with every claim. Google AI Overviews links cited sources in the overview text. ChatGPT includes sources as numbered references. The format differs, but in all cases, being cited with accurate context increases the chance a user clicks through to your site.
You cannot directly control how the AI phrases its summary. But you can control whether your page delivers a specific, trustworthy answer that the model treats as an authoritative source. Specific data and named sources make your page more likely to be treated as a primary citation rather than background context.
What schema does and does not do
Schema markup gets recommended in almost every AI search guide. The reality is more nuanced.
Industry research suggests that adding schema markup does not reliably improve AI citations. Pages with perfect structured data get cited at roughly the same rate as pages with none. Google had already removed FAQ and HowTo rich results from its search results earlier in 2026, removing the visible reward for schema implementation.
What schema does: it helps crawlers understand what type of content your page contains. FAQPage schema tells a crawler that a section contains question and answer pairs. Article schema tells a crawler this is editorial content. That labeling helps with indexing and parsing.
What schema does not do: it does not add information to your page. If your page contains nothing worth quoting, labeling it with schema does not change that. The model reads your page, finds nothing useful, and moves on.
The practical takeaway: add schema because it is good hygiene. Use Article or BlogPosting schema on your content pages, FAQPage schema on your question and answer sections, and Organization schema on your homepage. But do not expect schema alone to improve your citation rate. Focus on making your content extractable first.
What to fix first
If you want to appear in AI overviews, here is the priority order based on how the pipeline works.
First, fix crawlability. Check your robots.txt. Make sure GPTBot, PerplexityBot, ClaudeBot, and Google-Extended are allowed. This is a one-line fix per crawler and it is the most common gap.
Second, make sure you are ranking for the queries you want to be cited for. AI search engines pull from the top of their indexes. If you are not in the top organic results, you are not in the candidate pool.
Third, restructure your content for extraction. For every question your page should answer, put the answer in the first sentence under a heading that mirrors the question. Move background and context below the answer. Include specific data, named sources, and concrete examples. Keep paragraphs short. Use one H1 per page.
Fourth, add schema. Article schema for editorial content. FAQPage schema for question and answer sections. Organization schema for your homepage. Match the schema text to the visible text on the page.
Fifth, measure whether it worked. Check your AI visibility across ChatGPT, Perplexity, and Google AI Overviews. If you optimized inputs but never checked outcomes, you are guessing.
How to check where you stand
You cannot fix what you have not measured. Here is how to check your AI visibility.
Go to ChatGPT, Perplexity, and Google. Ask questions your customers would ask, using their exact phrasing. See whether your site appears as a cited source. This is manual, but it gives you a direct signal.
Check your server logs for AI crawler activity. If you see zero requests from GPTBot, PerplexityBot, or ClaudeBot over several weeks, either your site is not being discovered or your robots.txt is blocking them.
Run an audit. The Parceit audit engine checks your crawlability, your structured data, your content structure, and every other signal that determines AI search visibility. It tells you exactly what is working and what to fix first, in priority order.
Run a free audit at parceit.com/audit. You see your full score across every major AI search engine and a specific list of what to fix first. Takes under 60 seconds.
Frequently asked questions
How do I get my site to appear in AI overviews?
Make sure AI crawlers can reach your site through robots.txt, rank well enough in organic results to enter the candidate pool, structure your content so answers are easy to extract, and add schema markup as a hygiene factor. The combination of crawlability, ranking, and extractable content is what gets you cited.
Does adding schema markup guarantee AI citations?
No. Industry research suggests that schema markup does not reliably improve citation rates. Schema helps crawlers understand your content type, but it does not add information to your page. If your content has nothing worth quoting, schema does not fix that.
Why does my page rank well but not get cited in AI overviews?
This is an extraction problem, not a ranking problem. Your page was retrieved, but the AI model found it easier to extract a clear answer from a different page. Common causes: the answer is buried under introductory content, the page speaks in generalities, the HTML structure is hard to parse, or the page contains nothing unique compared to other retrieved pages.
How long does it take to appear in AI overviews after fixing crawlability?
AI crawlers re-read robots.txt on their own schedule, typically every 24 hours for sites they already know about. If your site was previously blocked, it may take days to weeks for the crawler to rediscover you and begin crawling your content. The sooner you fix the block, the sooner the process starts.
What is the difference between directional and decision-grade AI visibility measurement?
The IAB framework distinguishes between directional measurement, which checks whether you have the inputs that might help (like schema and crawlability), and decision-grade measurement, which checks the actual outcomes: are you present, prominent, and correctly portrayed in AI answers. Schema checking is directional. Citation tracking is decision-grade. Parceit's audit aligns with the decision-grade standard.