Ethan Liu | SEO & AI Search

Building Convertos + NudgeScout. Sharing technical SEO checks, AI search experiments & lessons from shipping both. Project evidence in the pinned post.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

I'm Ethan Liu. I build:

https://t.co/t1PIdH44lk - SEO audits & AI visibility.
https://t.co/6QLPrewGuR - website research & outreach.

Follow for technical checks, AI search experiments & lessons from building both.

Project evidence below. More: https://t.co/lvijsLxmRb

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

A blank starting rank → position 23 isn't a measured gain of 77 places. That calculation quietly assumes the starting position was 100.

In the 2025 example in our NudgeScout guide, I'd keep newly recorded rankings separate from keywords with two known positions. Otherwise the average mixes observations with guesses.

I build NudgeScout; the worked example is here:
https://t.co/T6QcU5xaxZ

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

For an ecommerce site, I'd start with suppliers or partners who already list stockists, then check whether your store actually qualifies. For editorial outreach, offer something useful to the specific article—a comparison, original photos, or instructions—not just a homepage link. Which site are you working on, and what have you tried so far?

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@abhijitnr I'd give the data-deletion date separately from the trial-end date. 'You lose editing access Friday, but can export until X' is much clearer than 'your trial expires.' Also check that the export link still works after the account drops to the free tier.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

Start with one job your SaaS replaces, then look for people describing how they do that job today. A specific workaround—copying the same data into three tools, for example—is a better reason to contact someone than 'founder' in their bio. Ask about that workaround first; if there isn't a real fit, don't pitch.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

You can't separate them from a before/after move where both changed. I'd want the same message tested across providers, then a format test within one provider, with comparable audiences assigned to separate groups. Count qualified replies or booked meetings per assigned prospect—not opens—and never send both versions to the same person.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

I'd split those 693 visits by device before judging the landing page. A phone visitor may still own a Windows PC, so a 'send the download link to myself' option could help. Then compare sources by Windows downloads and first completed recordings, not just visits—you need to know where people actually get a usable demo.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@madsmadsdk Moving the goal later is worth trying, but I'd judge it by first completed recordings, not just people passing screen one. I'd test the framing step separately if you can; changing both together makes it harder to tell what helped or where the drop-off moved.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@yildirimsertbas Google says that report includes links found over time, even if they've since disappeared, and groups target pages by canonical URL. I'd open one listed source/target pair before treating it as a new bug.
https://t.co/MOhKRszERy

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@joelsstafford @Guronnimo Are you counting inference per request or per finished task? Retries, failed runs and support still cost you. I'd work out that floor first, then test a price for a result customers care about, rather than let the model's token bill set the whole price.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@liebertjohanm I'd make the paid part clear in the post or store screenshots before changing the price. If people expect free, a cheaper surprise is still a surprise. Then compare new reviews and subscriptions from that traffic before deciding the price itself is the problem.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@Tajine1x @MilaChervenkova An EMD can still rank; that isn't a guaranteed keyword-domain boost. Google's guide says it limits excess credit from exact-match domains. I wouldn't assume a CTR lift without query-level data.
https://t.co/nfEhHJEbRV

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@realharleychu Was the higher-plan customer from today's PH launch, or an earlier signup? I'd keep those groups separate and follow the new PH users through their first return visit; today's total of one new payer doesn't tell you how that launch group converts yet.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

I’d mention the specific issue briefly, then ask whether it’s still a problem. An old review tells you what frustrated someone then, not whether they want to switch now. If the public profile doesn’t clearly identify the reviewer, use it for research rather than guessing who to contact.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@aadits25 @karthiknish A person copying one transcript and a developer embedding transcription in a workflow need different demos. I’d pick one for the next outreach batch and show the complete job, not just the API. Ask what they use now and what would make switching worth the setup.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

Has Googlebot fetched the new subpage yet? I’d compare URL Inspection’s last crawl time with the indexing-service submission time. That separates a discovery/crawl delay from a page Google has already crawled but not indexed; two hours alone doesn’t establish an update-wide slowdown.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@LarsMaehler @neilpatel I’d ask for the survey questions first. The post describes what 500 marketers think matters, not a test of what caused AI citations. An earned-versus-placed split would need definitions and source-level evidence; those percentages alone can’t tell us.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

The 10 free recipes give you a useful test before changing the price: can a new visitor pick one, run it, and get the promised result? I’d track that separately from pageviews. If people finish a free recipe but never need another, that points to a different problem than nobody trying one.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@Iroro_chad AlternativeTo looks relevant for Escourtly: it already groups Supademo with other demo builders. Show what your HTML capture lets buyers do differently, plus a working demo. That gives people a reason to compare, beyond another directory listing.
https://t.co/qh2EochMc3

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@mmalinowski83 @emaann28 Bing fetching it doesn’t tell you what Google received. Copy the exact submitted sitemap URL into GSC’s URL Inspection and run Test live URL. Under Page availability, check “Crawl allowed?” and “Page fetch”; a failure there gives you something specific to fix.

[Ethan Liu | SEO & AI Search @HeyEthanLiu]

@mohinimaheshwa For a 10,000-page site, the export needs a caveat: GSC’s example lists are capped at 1,000 URLs and aren’t a complete inventory. A URL missing from the indexed sample isn’t proof it’s unindexed. Keep the sitemap URL list separate and label unknown status before planning links.