Laura Iancu, the independent SEO behind Searchpedia, has had the expensive AI-visibility data through a big client, and she still doesn’t trust it on its own. On Unscripted SEO she explained why: she kept the synthetic numbers she pulled months ago, lined them up against what Google, Bing and Clarity report now, and found them “very different.” This SOP turns her habit into a check you can run on any client before you sell them an AI-search plan.
Get this SOP as a one-page PDF
The printable SOP – plus every other SOP, playbook and episode pack – is free inside The Vault. Create a free account and take it with you.
Open The Vault →Heard on the full interview · Unscripted SEO
Objective
You finish with a side-by-side sheet: what your paid AI-visibility tool says about a brand, next to what the brand’s own first-party reports say for the same period. Where the two agree, you can act. Where they don’t, you tell the client the number is modelled, not measured, and you hold the recommendation until your own data backs it. Laura’s rule is simple: use the tools, but “take them with a pinch of salt, especially because I compare the data.”
Key Steps
- Snapshot what the paid tool says, and date it. Export the brand’s AI-visibility numbers from whatever tool you pay for (Laura used Ahrefs Brand Radar) and save the file with the date in the name. The snapshot is the whole point. Laura could compare because she kept her data from October and November of last year.
- Pull the first-party side for the same period. Open the reports the brand owns: Google Search Console, Bing Webmaster Tools and Microsoft Clarity. Laura points to all three. Take whatever AI-surface reporting each one shows for the property, plus referral traffic from AI assistants in Clarity or GA4. (Which exact report you use in each tool is our addition; the reporting keeps changing, so take what the property actually shows today.)
- Script the pulls instead of buying another extension. Laura’s point: you don’t need “the newest, shiniest SEO tool extension” for synthetic data when the free tools already hold what you need – “all you have to do is prompt them properly and get the proper scripts.” Ask an AI assistant to write the export script for you. You don’t have to learn Python to do it.
- Line the two up topic by topic. Put the tool’s numbers and the first-party numbers in the same sheet, same date range, grouped by topic or product line. Mark each row agree, disagree or no first-party data. (inferred – Laura describes the comparison, not the sheet layout)
- Check the answer engines yourself. Jeremy’s addition from the same conversation: run the brand’s key questions in logged-out or anonymous sessions of ChatGPT, Perplexity and Claude. What they pull from the site in a clean session is close to what most people will see, and it updates fast when you change the page.
- Test before you advise. If a tactic only shows up in the tool, or only worked for someone else, test it on this client first. Laura tells clients “let me get back to you in a month. I need the time to test it myself” – because what worked for an expert, at their scale, “doesn’t mean it’s going to work for my clients.”
- Report the gap out loud. When the tool and the first-party data disagree, say so in the client report, and label tool numbers as modelled. Laura: “if I don’t, I’m just transparent and tell them as it is.”
- Re-run the comparison on a schedule. Keep every dated snapshot and repeat steps 1-4 each quarter. (inferred – the cadence is ours; the value of keeping old snapshots is Laura’s)
Cautionary Notes
- Synthetic is not measured. Tool visibility data is synthetic – modelled, not measured. Laura’s comparison found it “very different” from what the platforms report.
- The samples are small. Laura’s warning: nobody is rushing to test on really big websites, so most published results come from small samples.
- Quick wins crash. A tactic that moves an AI answer within the hour is easy to abuse, and easy to penalise. Laura: “That’s just riding a current wave that’s literally going to crash.”
- Don’t price small clients into big-tool data. Laura points out small clients won’t pay that kind of money for synthetic data. The free first-party reports come first.
- Direct traffic is not all people. Jeremy’s note from the episode: a lot of “direct” in GA4 is bots, and Clarity shows more of them than Google does. Check before you count it.
Tips for Efficiency
- Load Microsoft Clarity through Google Tag Manager so every client has it by default (Jeremy’s tip on the episode).
- Connect GTM and GA4 to Claude Code and ask it to “display my GA4 data in the style matching GA3” – then compare Clarity against GA4 in the same view (Jeremy).
- Name every export
brand_tool_YYYY-MM-DD. The date is what makes the next comparison possible. - Follow practitioners who publish tested work – Laura names Aleyda Solis – and still test it on your own client before you repeat it.
Sources & Relevant Episodes
- Laura Iancu, Searchpedia – calls herself a “search activist”, builds “sustainably and ethically”, and treats every AI-search claim as something to test on her own clients before she advises on it.
- Full interview recap: AI search quick wins with Laura Iancu
- Episode: Laura Iancu on Unscripted SEO · watch on YouTube