Month Three: 8,950 Impressions, 5 Clicks, and 2 Leads

Sixty percent of our Google impressions come from four pages that have produced three clicks in ninety days.

That is the number that reframed month three. Months one and two both led with impression growth as the proof the strategy was working. Month three is where I pulled the page-level export and found out what that growth was actually made of.

TL;DR: Google impressions went flat at 8,950 and clicks fell from 9 to 5. Sixty percent of those impressions come from four pages stuck at position 52 to 79. Bing AI citations dipped 20%, then tripled to 69 per day once cited-page breadth went from 4 to 13. Two organic leads arrived on 5 Google clicks, both through the payments quote form, from verticals holding under 1% of our impressions.

This is the third post in the ClickWerxs SEO build in public series. Real numbers, honest assessments, decisions reported against what the data actually did. Month one is here. Month two is here.

Reporting window: June 10 to July 9, 2026, with a separate section on July 10 to 23. Month two promised this report on July 9. It is late, and the two weeks of delay turned out to contain the most important data in it.

What month two committed to, and what we shipped

Month two closed with three decisions. We executed two of them.

CRM cluster launches. SHIPPED. Merchant services virtual terminal, CRM for small business, signs you've outgrown your CRM, and spreadsheet vs CRM all went live, pointing at the CRM features page and CRM onboarding, which were drawing impressions with no blog content behind them.

Hold title reworks until the Bing recrawl completed. HELD. No title-level changes were made to existing posts during month three. What that bought is in the Bing section below.

A focused Stripe alternatives post. NOT SHIPPED. The slot went to the AI SEO cluster. What happened to the Stripe alternatives page without it is further down, and it complicates that call rather than excusing it.

Where pages landed: four English blog posts now hold an average position inside the top 20 over the trailing 90 days. Gift cards for small business at 5.92, the Square Issuing V2 review at 8.17, the Square Issuing V2 gift card review at 10.82, and credit card processing for construction contractors at 19.77. Those are averaged positions on 55 to 81 impressions each, which is a thin sample, and none of them sits in the merchant services cluster the content strategy was built to win.

What we published in month three

Eighteen posts between June 10 and July 9, a pace of 4.2 per week. Month two ran 4.9 per week.

The composition changed more than the pace did. Nine of the eighteen were AI SEO posts: generative engine optimization, answer engine optimization, how to get cited by ChatGPT and Perplexity, E-E-A-T in 2026, and how to rank in Google AI Overviews, among others.

That was not the month two plan. Month two committed to a focused Stripe alternatives post as decision number one. We did not write it. The AI SEO cluster took its slot because Bing citations were the only channel showing compounding returns, and I would rather feed a channel that is moving than a keyword that is not. The Bing section below says that call was defensible. The Google section says it had a cost.

What the Google data shows, and what it hides

Month three (June 10 to July 9): 8,950 impressions, 5 clicks, 0.06% CTR, average position 63.6.

Month two (May 10 to June 9): 8,763 impressions, 9 clicks, 0.10% CTR, average position 66.2.

Impressions grew 2%. Clicks fell from 9 to 5. Average position improved 2.6 places and produced fewer clicks than the month before. Month two's post reported 8,223 impressions for that window; this export shows 8,763, because Search Console keeps attributing data for days after a window closes. Every figure here comes from a single export dated July 25, 2026.

Zooming out to the whole project, April 9 through July 9: 19 Google clicks on 21.6K impressions at an average position of 61.5. That is the cumulative result of three months and roughly 60 posts.

Nineteen clicks. That is the honest headline number for the quarter, and it is the one I would lead with if this were a report to a client rather than about my own site.

Now the part that matters. Across the trailing 90 days the site recorded 22 clicks on 24,231 impressions.

One detail before the table, because it is the kind of thing this whole post is about. Search Console's Pages report totals 25,498 impressions for that same window, not 24,231. Neither number is wrong. The site-level chart counts one impression per query where any of your pages appeared; the Pages report counts one per page shown, so a query that surfaces two of your pages registers twice. Comparing a page against a site-level total quietly understates its share. The percentages below use the page-level total, which is the only correct denominator for a page-to-page comparison.

Four pages account for 15,376 of those 25,498 page impressions, 60.3% of everything, and 3 of the 22 clicks:

Position 73 to 79 is page eight. Nobody scrolls to page eight. Those impressions are a record of Google matching a query to a page and then burying it, and they will convert at approximately zero regardless of how large the number gets.

The query data says the same thing from the other direction. Crypto queries drew 5,574 impressions at an average position of 75.2 and produced one click. POS and terminal queries drew 4,190 impressions at 72.7 and produced zero. Interchange queries drew 3,506 impressions at 79.5 and produced zero. Those three clusters are 68% of all query impressions and one click between them.

Meanwhile all 30 English blog posts combined account for 6.5% of impressions and 2 clicks.

I have spent two months reporting impression growth as evidence. The page-level data says the growth was concentrated almost entirely in service pages that are not close to ranking, on high-volume head terms a three-month-old domain has no business competing for yet.

The one genuinely encouraging Google number is the homepage: 265 impressions, 7 clicks, 2.64% CTR, average position 8.97. Brand search works. Everything else is still queuing.

The Stripe cluster deserves a note because we skipped the post and it improved anyway. "Stripe alternatives" ran 252 impressions at position 67.9 in month two. Over the trailing 90 days it sits at 750 impressions and position 48.0. Twenty positions gained with no new content pointed at it, and still zero clicks, because position 48 is page five.

The Bing AI numbers: a dip, then a step change

Month three (June 10 to July 9): 855 citations, 28.5 per day, 7.1 cited pages per day.

Month two (May 10 to June 9): 1,101 citations, 35.5 per day, 4.0 cited pages per day.

Citation volume fell 20% month over month. That looks like a loss. The second metric says it is not.

Average cited pages per day went from 4.0 to 7.1, a 78% increase. In month two the citations were concentrated on a handful of pages. In month three Bing was pulling from nearly twice as many distinct URLs per day. The base broadened while the frequency dipped, which is what re-anchoring after the May title rework was always going to look like.

Per Search Engine Land's coverage of the report's February 10, 2026 launch, average cited pages is defined as "the daily average number of unique URLs from a site referenced across AI experiences." It is a breadth measure, and breadth is the harder thing to build.

Cumulative across April 9 to July 9, Bing reports 2.6K citations at an average of 5 cited pages per day.

The two lines on that chart are the whole argument. Purple is citation volume, blue is how many distinct pages got cited. Blue climbs faster and climbs first. Every rise in the purple line through late June and July comes after the blue line has already moved.

The two weeks this report was late

Month two promised these numbers on July 9. What happened between July 10 and July 23 is the reason the delay was worth it.

The trailing seven-day citation average on July 9 was 45.6 per day. By July 23 it was 69.3. Average cited pages per day went from 10.6 to 13.0. July 22 set a record at 96 citations, and July 21 set a separate record at 17 distinct pages cited in a single day. For July 1 through 23 overall: 1,318 citations at 57.3 per day across an average of 11.8 pages, above every month one and month two figure on both volume and breadth.

Publishing slowed during this stretch. Six posts between July 12 and July 25, a pace of 2.6 per week against month three's 4.2. Citations more than doubled while output dropped 38%. The most plausible read is that June's AI SEO cluster finished indexing during July and the citation engine is now drawing on a far wider content base. I cannot prove that with the data Bing exposes. I can say the timing lines up and the breadth metric moved first.

Bing organic search moved in the same direction on a smaller base: 8 clicks on 477 impressions in month three, against 2 clicks on 143 impressions in month two. Across June 10 to July 23 the totals are 16 clicks on 1,069 impressions.

Cumulatively, April 9 to July 9, Bing organic sent 10 clicks on 629 impressions at a 1.59% CTR. Google sent 19 clicks on 21.6K impressions at 0.1% over the identical window. Bing showed this site to 3% as many people and converted them at roughly sixteen times the rate, because the impressions it did serve were not buried on page eight.

Two leads

Two organic leads in month three, up from one in month two.

The first is an ecommerce business processing roughly $25,000 a month. The second is a homebuilder and renovation contractor at around $250,000 a month. A tenfold spread in volume, two unrelated verticals, both inbound from organic search inside the same thirty days.

Those two lines matter more than any number above them, and not because two is a large figure.

Neither lead came from the verticals that generate our impressions. Crypto, POS, and interchange queries are 68% of all query impressions and produced one click and zero leads. Construction and contractor content, by contrast, accounts for roughly 0.5% of this site's impressions. It also produced the only English blog post that has earned a click and cracked the top 20: credit card processing for construction contractors, position 19.77 on 81 impressions.

Both leads arrived through the payment processing quote form. Not the CRM demo request. Not a web design enquiry. Nothing attached to the AI SEO cluster that took nine of month three's eighteen post slots.

That form captures no landing page and no referrer, so I cannot prove which page produced either lead, and closing that gap is a month four job. What the data does support is that the impression distribution and the lead distribution are close to inverted. The four pages generating 60% of our visibility have generated nothing. A vertical generating half a percent of our visibility generated the largest lead the site has produced.

The measurement problem is worth stating plainly, because the arithmetic is uncomfortable. Google sent 5 organic clicks in the month three window. Two leads cannot plausibly come out of 5 clicks. Whatever produced them arrived through a path Google Analytics attributes poorly or not at all: Bing organic, an AI assistant answer that led to a later branded search, or direct entry after the brand was seen in a cited answer.

For context on the funnel, GA4 recorded 1,941 page views, 1,466 users, and 1,609 sessions between April 26 and July 24. Of those users, 1,465 were first-time visitors. The site is running on almost entirely new traffic. Across the same 90-day period there were 22 form starts and 11 generate_lead events across all channels, of which the two discussed here are the organic-attributed subset for month three.

That chart covers April 9 to July 9, a slightly different window than the figures above, which come from the April 26 to July 24 export. What it shows either way is spike-driven traffic rather than a rising baseline. The peaks are individual posts getting discovered. The floor between them barely moves, and the floor is what compounding traffic actually looks like.

The counterargument, and it got stronger this month

The case against this whole approach: citation counts are a vanity metric with no proven path to revenue, and month three is the month that argument got its best evidence.

Search Engine Land's coverage of the AI Performance report states the limitation directly. The metrics "only reflect citation frequency" and "don't indicate ranking prominence, how pages contributed to specific answers, or whether citations translate into actual traffic or business value." That is Microsoft's own reporting tool, described by the trade press covering its launch, conceding it cannot connect citations to outcomes.

So the skeptical read of month three is available and defensible: Google clicks fell 44%, three of the four largest pages produce nothing, the one post we committed to writing never got written, and the channel we are celebrating is the one nobody can attribute revenue to. Sharper still, half of month three's output went to the AI SEO cluster and both leads came through the payments quote form. The content that produced citations and the content that produced revenue were not the same content.

I still think the allocation is right, and month four is where that gets tested rather than argued. A domain this young cannot win head terms in Google on a three-month timeline. It can get cited today. The bet is that citation breadth compounds into brand recognition and branded search, and the homepage's 2.64% CTR at position 8.97 is the first weak signal that it might.

If month four produces citation growth with no movement in branded search or leads, the bet is wrong and I will say so here.

What changes in month four

Kill the head-term service pages as an impression source. Crypto, POS, and interchange are 68% of query impressions and one click. No further content gets pointed at those head terms. The volume is real and the intent is wrong for a domain at this age.

Rewrite the four buried pages for long-tail intent instead of category terms. Position 79 on "interchange fees" is worth less than position 12 on "amex interchange rates for small business." The 90-day query export has 1,047 rows and the long tail inside it has not been worked at all.

Follow the leads, not the impressions. Both month three leads came from ecommerce and residential construction, both through the payments quote form, and both verticals sit under 1% of our impressions. The industry content gets the next cluster: contractors, homebuilders, renovation, and ecommerce operators at the volumes these two leads represent. That is a smaller total addressable search volume than crypto and it is the only volume that has converted.

Ship the Stripe post that month two promised. Position 48.0 on 750 impressions, achieved with no dedicated content, is the clearest unexploited gap in the data. Skipping it twice would be a pattern rather than a tradeoff.

Instrument the lead path properly. Two leads that cannot be traced to a channel is a reporting failure, not just an attribution limitation. Month four adds a "how did you hear about us" field and landing-page capture to the contact form. Both are crude instruments and both beat what is running now.

I am not going to forecast what any of that produces. Three months of data on this domain is not enough to predict a fourth, and publishing a number I have no basis for would be the least honest thing in an otherwise honest report. Month four reports these four decisions against what the data does, including the ones that do not get shipped.

The next report goes up on August 25.

Frequently Asked Questions

Should I stop tracking Google impressions if most of them come from deep-ranked pages?

No, but track them segmented by page and position band rather than as a site total. A site total blends pages at position 8 with pages at position 79, and the second group can grow indefinitely without producing a single click. The useful cut is impressions from pages ranking inside position 30, which is where clicks actually start. On this site that segment is a small fraction of the 25,498 page-level impressions and it is the only part of the number that forecasts traffic. Check which report you are reading before you compare anything: the Pages report and the site-level chart return different totals for the same window, because one counts an impression per page shown and the other counts one per query.

Why did Search Console show different numbers for the same window two months apart?

Search Console continues attributing data for several days after a date passes, so any figure pulled on the final day of a reporting window understates the total. Month two's post reported 8,223 impressions for May 10 to June 9. The same window in a July 25 export shows 8,763. For monthly reporting, pull the export at least three days after the window closes, or expect to revise.

How many AI citations does a page need before it produces measurable traffic?

There is no published threshold, and Bing's own report cannot answer it. The AI Performance report measures citation frequency and does not connect citations to sessions or conversions. The practical approach is to track citation volume and branded search volume as two separate series and watch whether the second one moves after the first does. That relationship, if it exists, is the entire case for optimizing toward AI citation, and it has not been demonstrated on this site yet.

Is average cited pages a better metric than total citations?

For a site building AI visibility from zero, yes. Total citations can be produced by a single page getting repeatedly referenced, which is fragile and disappears if that page's signals change. Average cited pages measures how much of your content the AI considers citable. On this site the count went from 4.0 per day in month two to 13.0 per day by late July, and the volume increase followed the breadth increase rather than the other way around.

My highest-volume keywords and my actual customers are in different verticals. Which do I write for?

Write for the vertical that converts, and treat the high-volume cluster as a long-term asset you are not ready to collect on. Search volume is only worth pursuing when your domain can realistically rank inside the top 20 for it, and a young site usually cannot on head terms. On this site the crypto, POS, and interchange clusters carry 68% of impressions and have produced zero leads, while construction content carrying 0.5% of impressions sits alongside the largest lead the site has generated. The decision rule is to rank where you can convert first and buy the volume terms later with the authority you earn.

The two leads in this report came from an ecommerce operator and a residential builder, out of a vertical that holds under 1% of our search impressions. If you are building SEO infrastructure for a service business or a payments company and want the reporting done at this level of honesty, ClickWerxs builds and manages it.

Operator opinion. Not a guarantee of results for your site. SEO timelines vary by domain age, niche, and content quality. All first-party figures come from Google Search Console, Bing Webmaster Tools, and GA4 exports dated July 25, 2026.

Kaleb Dickhaut — Founder, ClickWerxs Kaleb built ClickWerxs from the ground up, from payment processing ISO to the Command Center platform to the AI SEO methodology this blog runs on. He has onboarded hundreds of small businesses onto payment and CRM systems. linkedin.com/in/kaleb-dickhaut

Sources

1. ClickWerxs blog, first-party data — Search Console, Bing Webmaster and GA4 figures in this post are measured from ClickWerxs's own properties over the stated period. They are a past result for one domain in one vertical and are not a projection of what any other site will achieve. Operator data.
2. Competitor and platform pricing referenced in this post is taken from each company's published pricing or legal pages as of the date noted in text, and is subject to change without notice. ClickWerxs does not link to competitor websites; references are given in text so they remain verifiable. No affiliation is implied.

ClickWerxs sells SEO and AI visibility services and earns revenue from those engagements. First-party figures are past results for this blog and are not a promise of future performance. This is operator opinion, not professional advice.