What Keywords Does My Site Rank For? A 30-Minute Audit Across SEO and AI Search

In this article, you’ll run a four-method audit that pulls every keyword your site ranks for in Google, every prompt your brand surfaces in across ChatGPT, Claude, Perplexity, and Gemini, and every gap your competitors are quietly winning while you’re not looking. You’ll leave with a single ranked list of pages to fix this week, the queries that are about to break onto page one, and a clear view of where your AI search presence is leaking traffic.

Why finding your ranking keywords matters

You can’t fix what you can’t see. Your ranking list is the cheapest map of opportunity you’ll ever own.

Pages stuck on positions 8 to 20 are your fastest wins. The hard part of ranking is already done. A title rewrite, a few internal links, and a refresh on a page that already ranks #11 will outpull launching a brand new page nine times out of ten.

The same is true for AI search. If your brand is mentioned in 30% of “best CRM for startups” responses on ChatGPT but 0% on Claude, you don’t have a discovery problem. You have an engine-mix problem, and that’s a faster fix than starting from zero.

The audit below gives you both views in under thirty minutes.

Method 1: Google Search Console

Google Search Console is the only place that shows you the actual queries Google sent traffic for. No estimates. No third-party crawl. The data comes straight from the source, and it’s free.

Open Search Console, pick your property, click Performance → Search results.

The default chart shows clicks, impressions, average CTR, and average position over the last three months. Switch on all four metric tiles. Then click the Queries tab below the chart.

You’re looking at every Google query that earned you at least one impression in the period. That’s your real ranking universe.

What to do once you’re staring at the list

Sort by Position ascending and look for keywords sitting in positions 8 to 20 with meaningful impressions. These are the quick wins. Export the table and tag them.

Sort by Impressions descending while filtering for CTR < 2%. You’ve found the queries where Google ranks you, but no one clicks. The cause is almost always one of three things. A weak title tag, a meta description that doesn’t match intent, or content that’s ranking on accident for a query it doesn’t really serve.

Filter to Position 1 to 3 and sort by Clicks ascending. Anything ranking #1 with under 50 clicks a month for a meaningful query points to a search volume mismatch. The volume estimate you trusted from a third-party tool was wrong. Trust GSC instead.

Use the URL filter to triage by page

Switch to the Pages tab, click any URL, then go back to Queries. You’re now looking at every keyword that one specific page ranks for. This is the single most useful view in GSC for content refreshes. Page by page, you see the long tail your URL accidentally captured. Every query the page wasn’t written for but ranks for anyway is a bonus topic to expand into.

Where GSC stops being useful

Search Console only shows Google. It hides queries below an anonymity threshold (your “hidden queries”). It caps history at 16 months. And it can’t tell you a single thing about Bing, ChatGPT, Claude, or Perplexity. For everything outside Google, keep going.

Method 2: Google Analytics 4

GA4 doesn’t store keyword data on its own. What it does, once you connect it to GSC, is pin every keyword to a business outcome.

In GA4, go to Admin → Property settings → Product links → Search Console links and connect your verified GSC property.

Once linked, head to Reports → Acquisition → Search Console → Queries. You now see every Google keyword next to engaged sessions, conversions, and revenue.

Sort by Conversions descending. The keywords near the top are the ones you protect with your life. Defending #2 on a money keyword is worth more than launching three new posts on traffic-only topics. Sort by Engagement rate ascending. The keywords at the top are the ones where you’re catching the wrong audience. Either rewrite for the actual intent or stop trying to rank.

What this view adds that GSC alone can’t

GSC tells you which keywords drive clicks. GA4 tells you which clicks drive money. The difference matters because most “high-traffic” keywords in B2B are noise. The keyword that brings 80 visits a month and converts at 6% beats the one that brings 3,000 visits and converts at 0.05%, every time.

Use this view to build the keyword strategy document your team actually argues over. Which queries you’re going to win, which you’re going to defend, and which you’re going to abandon.

Method 3: Ahrefs and Semrush

GSC shows you what you rank for. Ahrefs and Semrush show you what you could rank for and what your competitors already do.

Ahrefs Site Explorer

Drop your domain into Site Explorer and open Organic keywords. Ahrefs estimates every keyword you rank for in the top 100, with position, volume, KD, and traffic.

The filter combination that surfaces quick wins:

• Position: 4 to 20

• Volume: 100+

• KD: ≤ your domain’s average

Sort by Traffic potential descending. The top of that list is your prioritized refresh queue.

Semrush Position Tracking

Semrush’s Position Tracking tool monitors a defined keyword set every day, by device and location. It’s the right tool when you’ve already chosen your target list and need to watch movement instead of discovering new keywords.

Keyword Gap (Semrush) and Content Gap (Ahrefs)

Both tools let you compare your domain against three or four competitors and surface keywords they rank for that you don’t.

Filter to Missing keywords with volume > 200 and KD ≤ 40. Every row is a content brief in waiting.

A note on free tools

If you’re not paying for Ahrefs or Semrush yet, the free SEO tools you can stitch together get you most of the way. The keyword rank checker handles spot-checks, the SERP checker shows who’s on page one before you commit to a topic, the keyword difficulty checker handles fast triage, and the keyword generator covers fresh ideas. Use the website authority checker to benchmark yourself against the competitors above you in the SERP.

A quick incognito Google search is fine for sanity-checking a single query. It’s a terrible system for tracking more than five.

Method 4: Analyze AI

Methods 1, 2, and 3 cover Google. The audit isn’t done.

Roughly one in four of your buyers now opens ChatGPT, Claude, or Perplexity instead of Google for at least some of their research. They’re asking “best CRM for B2B sales teams under 200 reps” and reading a synthesized answer that names three brands. If your brand is one of the three, that’s an organic acquisition you didn’t know happened. If it’s not, you’ll never see the lost session in Google Analytics. There’s no impression to count.

This is what Analyze AI tracks. Instead of a query → ranking position relationship, you’re looking at a prompt → mention/citation/sentiment relationship.### Step 1: Add the prompts your buyers actually ask

Inside Analyze AI, you build a prompt set that mirrors how buyers describe your category. The categories that matter:

• Best [category] for [use case]. “best project management tool for distributed engineering teams”

• Alternatives to [competitor]. “alternatives to Asana for product teams”

• [Brand A] vs [Brand B]. “Linear vs Jira for engineering”

• Top [category] for [segment]. “top CRMs for healthcare SaaS”

• How to [problem your product solves]. “how to track AI search visibility”

You don’t have to think them all up. The platform suggests prompts based on your domain, your competitors, and the patterns it sees in real prompt logs.

Every prompt runs across ChatGPT, Claude, Perplexity, Gemini, and Copilot daily. The platform records four things. Whether you appeared, where you appeared in the answer, how the answer framed you, and which sources the engine cited.

This is the AI search equivalent of GSC’s Performance report. Same job, different surface.

Step 2: Read the engine breakdown

Different engines pull from different places. Treating them as one channel is a mistake. The visibility you have on Claude tells you almost nothing about your visibility on ChatGPT.

In our analysis of 83,670 AI citations:

Here’s the takeaway. If your audience is engineering-heavy and uses Claude, your blog content matters disproportionately. If they’re researching enterprise software in ChatGPT, your product pages and your Wikipedia presence matter more. The same content investment will pay back differently per engine.

Step 3: Find the prompts where competitors win and you don’t

This is the AI search version of Semrush’s Keyword Gap.

Each row is a prompt where a competitor is recommended and you aren’t. Each one is a content gap, a citation gap, or a positioning gap. Sort by competitor mentions and read the actual answer text. You’re looking for two patterns:

1. Topical absence. The engine doesn’t recommend you because you don’t have a page on this use case. The fix is easy. Write the page.

2. Citation absence. The engine has heard of you but cites third-party sources that mention competitors and not you. The fix is harder. Earn coverage from the sources that get cited in your space.

Knowing which one you’re dealing with tells you whether to brief content or do digital PR.

Step 4: See which sources AI engines actually trust in your space

Most teams optimize their own pages. AI engines cite their own pages 13% to 22% of the time, depending on the engine. The other 78% to 87% of citations go to third parties.

Sort by citation count. The top of the list is your earned-media target list. If G2, Capterra, and a specific industry blog show up in 60% of answers in your category, those are the three properties to influence. Self-promotional content on your own blog will not move the needle on engines that don’t cite you in the first place.

Step 5: Watch how AI describes you, not just whether it mentions you

Visibility without context is half the picture. A 90% mention rate inside answers that frame you as “expensive and limited” is worse than a 40% rate inside answers that call you the safe enterprise choice.

The Perception map clusters the language AI engines use about you. Risk terms (“limited integrations,” “pricing controversy,” “outdated UI”) show up here before they show up in your sales calls. Treat them like NPS open-text feedback. Each one is a content brief or a product fix, not a vanity score.

Step 6: Tie it back to traffic and pipeline

Visibility only matters if it lands somewhere. Connect GA4 and the platform shows the AI sessions, the engines they came from, and the pages they landed on.

This is the column your CMO actually cares about. Not “visibility %”, but “AI sessions, conversion rate, and pipeline this month.” When you find a page receiving 800 sessions a month from Perplexity, you’ve found the topic to invest in next.

The Kylian AI case is a useful reference. Their AI traffic scaled from 200 to 1,000+ sessions per month with a 5% conversion rate, well above the typical blog benchmark of 1% to 2%. AI search traffic isn’t smaller than SEO traffic in a way that matters. It’s smaller in volume and higher in intent, which is exactly what you want for pipeline.

The 30-minute audit checklist

You don’t need to run all four methods every week. You need to run the right one at the right cadence. Here’s the schedule that works for most marketing teams:

The audit itself, end-to-end, takes around 30 minutes once your tools are connected. The Analyze AI side runs on its own and surfaces deltas in a weekly digest, so the cognitive cost of “remember to check” disappears.

How to turn the audit into a content plan

You now have four data sources. The trap is treating each one as a separate to-do list. Don’t. Merge them into one prioritized queue.

Refresh before you write

Pages on positions 8 to 20 (from GSC) that already convert (from GA4) get refreshed first. Add depth, fix the title, route three to five internal links from related articles, and ship.

Write into the gap

Competitor-only keywords (from Ahrefs/Semrush) and competitor-only prompts (from Analyze AI) get queued as new briefs. The crossover is gold. A keyword and a prompt where competitors win and you don’t is a brief that pays back in two channels at once.

The AI Content Optimizer takes any of those URLs and scores them for AI search readiness, then surfaces line-level fixes (entity coverage, citation gaps, structure issues).

Earn the citations that move the needle

Your “uncited gaps” list (from the Sources view) becomes a digital PR brief. Pitch the publications and review sites AI engines cite, not the ones that look good on a media plan. The list is in the data, not in your head.

Don’t link to the wrong tools

When you build internal links during a refresh, route to your own depth and to free tools that match the reader’s intent. A reader on a “how to find ranking keywords” page is one click from converting if you hand them a keyword rank checker inline. They’re three clicks from leaving if you send them to a vague pillar page first.

How to put the whole thing on autopilot

Running a 30-minute audit every week is fine. Running it every Monday at 7am with the result already on your desk is better.

This is where Analyze AI’s Agent Builder changes the math. Most teams know Analyze AI as a dashboard. Underneath sits a programmable substrate (180+ nodes, 34 pre-built data recipes, 13 input primitives) wired to your AI visibility data, your GA4, your GSC, and your competitor list. You can stand up agents that run the audit for you on a schedule.

A few audits worth automating today:

• Monday keyword brief. Schedule for Monday 7am. Pull GSC position 8–20 queries with rising impressions, pull AI visibility losers from the last 7 days, assemble into a single doc, email it to the content team. The “what should we work on this week” meeting stops existing.

• Competitor visibility diff. Schedule daily. Run the competitor comparison recipe and post to Slack if any competitor gained more than 10% visibility share on tracked prompts.

• Citation-stop alert. Schedule daily. Flag any of your pages that lost AI citations in the last 24 hours and Slack the owner with the page and the last five prompts that stopped citing it.

• Content refresh queue. Schedule weekly. Cross GSC position 8–20 queries with the AI Content Optimizer score. If score is below 70, draft an outline for refresh and drop it into Notion.

Webhook agents take it further. A new piece of negative AI coverage triggers a comms response brief. A closed-won deal triggers a case-study draft against the prompts that customer cared about. A CMS publish event triggers an internal-linking pass on the new article.

The audit you ran for thirty minutes today becomes a continuous radar tomorrow. That’s the unlock most teams miss when they think “AI search tool” and stop at a dashboard.

Wrapping up

Finding the keywords your site ranks for is no longer a Google-only job. The audit that matters in 2026 covers four surfaces:

• GSC for the truth about Google performance

• GA4 for which of those rankings convert

• Ahrefs or Semrush for what your competitors rank for and you don’t

• Analyze AI for the prompts your brand surfaces in across ChatGPT, Claude, Perplexity, and Gemini

Run them once and you have a list. Run them weekly and you have a system. Automate them and you have leverage. The brands compounding visibility right now aren’t the ones publishing more. They’re the ones who know exactly which page, which keyword, and which prompt to fix next, and they know it before they sit down on Monday.

See where your brand ranks across every AI engine and ship the audit on autopilot.