Your First SEO Audit

In a good quarter of an hour you get an honest picture of where your website stands on Google: what already finds you, where you narrowly miss the clicks, and what to do first. All by prompt, without clicking through Search Console and technical tools.

Most people don't know where their website really stands on Google. You suspect there should be "something with SEO", but the tools are clunky and the reports are full of jargon. An SEO audit clears that up: it's a status check. Where does your site show up in search, how well, and what is holding it back? By the end of this tutorial you have four things: an overview of what you're found for, a list of the queries with the fastest lever, a short technical check of your homepage, and a shortlist of the three most important actions, prioritized by impact.

What you need

• Search Console connected (required). Google Search Console is Google's free tool that shows which queries your site appears for and at what position. Without it, your agent has no real numbers. To connect it, see Connect Marketing Accounts to Claude. If you work in ChatGPT, the ChatGPT setup tutorial applies in the same way.
• DataForSEO connected (for the technical part). The load-time check in step 3 comes from DataForSEO, a connected SEO data source. Without it, you skip the load-time part in step 3: the indexing check still runs, because it comes from Search Console.
• GA4 optional. You don't need Google Analytics for this audit. It additionally shows what visitors do on the site (not how they find you). For the four steps here, you leave it aside.

A word on language: I write "your agent" whether you use Claude or ChatGPT. The prompts work in both.

1. Visibility overview: what are you found for?

First the big picture: which queries bring you to Google, which pages collect hits, and whether it's going up or down.

Give me for [workspace] a visibility overview from Google
Search over the last 28 days: the most important queries and
pages with clicks, impressions and position, compared to the
28 days before.

What you get back:

Time frame: 28 days (comparison: 28 days before)
Impressions   9,800   (before 9,700)   ~ same
Clicks           58   (before 71)      slightly down
Avg. position  28.6   (before 30.4)    marginally better

Top queries by clicks:
Query                        Clicks  Impr.  CTR    Avg. pos.
client-a                        18     28   64%    4.2
client-a driving school          9     15   60%    1.1
driving school graz              1     40    3%    1.2
driving license cost graz        0    190    0%    9.1
motorcycle license graz          0    150    0%   32.4

Three terms to take away: impressions means how often your page appeared in the results; position is the average rank (1 = at the very top); the CTR (click-through rate) tells you what share of impressions turns into clicks. The revealing part is in the top list: almost all clicks there come from people searching the company name (brand search). For the real topic searches like "driving license cost graz" there are many impressions but almost no clicks: sometimes because you sit just below the click zone (position 9), sometimes because you're stuck on page 3 (position 32). In short: those who know you find you. Those who search for the service don't yet.

✓ Checkpoint: You have a clicks, impressions, and position figure for the period plus the previous period, and you see which queries actually bring you clicks.

Doesn't work? If you get "no data" or an empty list, either Search Console isn't connected (see "What you need") or the property is still too new: Google needs a few days before data accumulates.

2. Almost-there chances: where is work worth it first?

Now the most rewarding part. Queries at position 5 to 15 sit at the bottom edge of page 1 or just behind it: close enough that a small push lifts them into the click zone, far enough that they bring nothing yet. In SEO this is called "striking distance".

Show me for [workspace] the queries where I sit at
position 5 to 15 and still get barely any clicks.
Sort by impressions, the ones with the most potential first.

What you get back:

Almost there (position 5-15, enough impressions, barely any clicks):
Query                        Impr.  Avg. pos.  Clicks
driving license cost graz     190      9.1       0
driving school graz prices    150      6.8       0
b197 license graz             110      9.7       0
intensive driving course       32     12.5       0

These four are already on page 1, at the bottom. The lever is small: on the page that ranks for "driving license cost graz", sharpen the title and expand the content. Because the ranking already exists, such a polish often works faster than a brand-new article for a topic you don't rank for at all yet.

✓ Checkpoint: You have a handful of queries at position 5-15 with decent impressions: those are your first areas to work on.

Doesn't work? If the list is empty, you may still rank further back for your topics. Then ask instead for position 15 to 30: that shows you where you already appear at all, just deeper down.

3. Technical quick check: found and fast enough?

Two basic technical questions for your most important page: does Google find it at all (indexing)? And does it load fast enough on mobile? Both in one prompt.

Check my homepage for [workspace] on two things:
Is it indexed by Google? And how fast does it load on
mobile (Core Web Vitals)? Explain the values to me in plain language.

What you get back:

Indexing (homepage):
  Status     indexed, listed on Google
  Canonical  clean (page points to itself)
  Last visited by Google: 1 day ago

Load time on mobile:
  Performance score      76 / 100
  Load time (LCP)        3.9 s   (target: under 2.5 s)
  Layout shifts (CLS)    0       (target: under 0.1; very good)

Indexed means: Google knows the page and can serve it. That's the basic prerequisite, and here it's met. LCP (Largest Contentful Paint) measures how long it takes until the largest visible area has loaded: here 3.9 seconds, Google wants under 2.5. That's too slow and a real point to work on. CLS measures whether the layout jumps around while loading (0 = nothing jumps, very good). Bottom line: the page is found, it just loads too sluggishly on mobile.

✓ Checkpoint: You know whether your homepage is indexed, and you have a load-time figure with the Google target value next to it.

Doesn't work? If the status reads "not indexed" instead of "indexed", that's a serious finding: then the page can't rank at all. Have your agent name the reason (often a noindex directive or a wrong canonical) and save it for the very top of your action list. If the load-time block is missing, DataForSEO isn't connected: then you keep the indexing part and skip the load time.

4. Action shortlist: what first?

Four steps, many numbers. Now you have the agent prioritize: not everything at once, but the three levers with the greatest impact.

Summarize the findings and give me the three most important levers
I should tackle first, sorted by impact.
Be honest about what works fast and what takes longer.

What you get back:

The 3 most important levers (by impact):

1. Strengthen "almost there" pages        [biggest lever]
   License/driving-school pages sit at pos. 5-10.
   Sharpen titles + expand content → into the click zone.
   Why first: the ranking is already there, you just lift it.

2. Make the homepage faster on mobile
   LCP 3.9 s → target under 2.5 s. Compress images,
   load the largest element first.

3. Tackle the snippet problem
   Pages with many impressions but almost 0 clicks:
   write more attractive page titles.

This order follows a simple logic: first lift what is already almost at the top (step 1), then the technical side (step 2), then the fine-tuning of the titles (step 3). That's your roadmap.

✓ Checkpoint: You have three concrete actions in a defined order, with a reason why the first comes first.

Doesn't work? If the answer stays vague ("improve content"), ask specifically: "Which page exactly, and what should the title say?" The agent has the data, it just needs to get more precise.

Variations

• "Which of my pages has many impressions but a poor click-through rate?" (targets the snippet problem directly)
• "Compare my homepage on mobile and desktop." (load-time problems often affect only mobile)
• "Which topics do I not rank for at all yet, even though they fit me?" (uses DataForSEO for ranking comparisons; skip without DataForSEO)

Honest limit

• No ranking promise, and SEO takes time. Changes take effect over weeks to months, not overnight. No one can guarantee you the number 1 spot; reputable advice never does. The audit shows you the direction, not a shortcut.
• The audit finds, it doesn't build. This tutorial diagnoses. Rewriting the titles, fixing the load time: that's the next step, carried out by you or your developers.
• Single-page check, not the whole website. You check indexing and load time here per page (the homepage in the example). That one page is clean doesn't mean all of them are. A true all-round scan needs its own, larger run.
• The avg. position is a mean. Position 9 can mean: sometimes rank 4, sometimes rank 15. And it mixes the normal blue-link rank with spots in map boxes (local pack), snippet boxes, and AI answers at the top of the results. Use it as a direction, not as an exact rank.
• Few impressions are noise. For a query with only a few impressions, a single position is chance, not signal. Concentrate on the rows with decent impressions.

What's next

You now know where you stand in classic Google search. But a growing share of searches no longer runs through Google at all: people ask ChatGPT and other AI assistants whether a company like yours exists. In the next tutorial you check whether and how you show up there. → Visible in ChatGPT: Check Your Company's AI Visibility