A giant a part of search engine optimisation isn’t technique. It’s the menial, repeatable repairs: re-running the weekly website audit, catching the publish that quietly shed a 3rd of its rankings, noticing the DR 80 backlink you misplaced final Tuesday, checking whether or not AI assistants nonetheless describe your product precisely.
All of this wants somebody within the room, reliably, on a schedule, doing the identical diligent checks each time. But it surely doesn’t want you notably. It simply must ping you when one thing’s value your consideration.
That is the a part of search engine optimisation you’ll be able to hand to Agent A.
Listed here are a few of the finest use circumstances you may get began with. We’ve already constructed most of them as abilities or apps in Agent A, however I’ve additionally given you a starter immediate to make use of when you’d prefer to construct them your self.
What’s Agent A?
Agent A is a advertising agent from Ahrefs—an AI assistant with direct entry to the complete Ahrefs dataset that may perform advertising duties autonomously, reasonably than simply reply questions.
Agent A contains:
- Unrestricted entry to Ahrefs endpoints. Each endpoint we use to construct Ahrefs is out there, together with many you can not attain through API or MCP.
- Severe tech stack beneath. Postgres for state, Flask for UIs, an OpenRouter proxy with 300+ fashions, net fetch with full-page parsing, PDFs, OCR, scheduled jobs.
- Native connectors to advertising instruments. Slack, HubSpot, GitHub, Notion, Linear, Mailchimp, Resend, SendGrid, Stripe, Gong, WordPress, Airtable, Apify, and even Semrush.
- Professional ability library. The Ahrefs group has contributed pre-built advertising abilities and purposes that encode how we really work.
Key phrase analysis isn’t exhausting. It’s tedious. You pull 4,000 candidates, then spend a day throwing 3,800 of them away, studying SERPs one key phrase at a time, and dragging the survivors into one thing that resembles a plan.
Sam, our VP Advertising and marketing, determined he was accomplished doing that by hand. He had Agent A construct a key phrase analysis device that can take your area of interest (e.g., “espresso”), kind them into clusters, and provide the capacity to spin up a content material temporary for every one.
While you kind a distinct segment into the device, Agent A expands it into seed key phrases, then does a whole afternoon’s work in twenty minutes.
It pulls actual quantity from Key phrases Explorer and reads the SERP for each surviving candidate, analyzing search intent, key phrase issue, and who’s already rating. For every key phrase, it counts the page-type make-up of the highest 10 (e.g., editorial articles versus product pages, YouTube movies, or Reddit threads).
Every thing comes out graded Go, Perhaps, or Skip, with the simple wins flagged, plus your rivals mined for hole key phrases you don’t cowl but.
Twenty minutes later, you come again to 5 tabs:
- Overview (with charts)
- A graded key phrase record
- Clusters grouped into subjects (every with pillar key phrase, whole quantity, site visitors potential, common issue, and SERP make-up at a look)
- Competitor Hole record
- Hub-and-Spoke Map of the way it all hyperlinks collectively
Then, when you click on on the “Get temporary” button on any key phrase, Agent A writes the spec a content material editor would: three title choices and a meta description, a URL slug, a word-count goal set to beat the pages at the moment rating, a full H1/H2/H3 define, FAQs mined from the precise Reddit and Quora threads in that SERP, internal-linking tables, and a differentiator aimed on the weakest competitor on web page one.
If the SERP is video-heavy, it tells you to shoot a YouTube video alongside the article.
We’ve constructed Sam’s Content material Key phrase Analysis app in Agent A, so all you must do is to put in it.
When you’d prefer to construct your individual, right here’s a starter immediate:
Starter immediate
Construct me a key phrase analysis device. Enter: a distinct segment (one phrase or phrase). Pipeline: (1) fetch key phrase options + world and top-country quantity from Key phrases Explorer; (2) vet every candidate, dropping low-volume, single-word, and off-intent phrases; (3) run a SERP evaluation per surviving key phrase to attain issue and classify intent; (4) cluster key phrases into subject teams; (5) optionally available: generate a content material temporary per cluster. Present clusters within the UI with quantity and intent per key phrase, and an “export” button. Run as background jobs the UI polls.
Website audits are nice at producing findings and horrible at getting them mounted. The report fills up with points and the problems sit there.
So right here’s the right way to automate that in Agent A.
Arrange a job that runs each Sunday. Ask Agent A to drag your most up-to-date Ahrefs Website Audit, rank the problems by precedence, and open a GitHub pull request scoped to the high-severity fixes: indexability issues, damaged pages, damaged inner hyperlinks, and extra.
The PR lands with a guidelines and the audit knowledge connected, so the developer choosing it up Monday morning doesn’t should re-derive what’s fallacious.
Each processes are already in-built Agent A as Expertise. The Website Audit Discovery ability surfaces and prioritizes the problems, and Website Audit Concern Fixer drafts the fixes.
As soon as launched, you’ll be able to ask Agent A to run this each Sunday so it’s automated.
Starter immediate
Construct me a weekly site-audit-to-GitHub job. Each Sunday, fetch my newest Ahrefs Website Audit, filter to high-severity points (damaged hyperlinks, lacking canonicals, redirect chains, indexability), group them by repair kind, and open one GitHub PR per group with a guidelines physique and the affected URLs. Skip any problem that’s already in an open PR. Put up a one-line abstract of what was opened.
Site visitors decay is invisible each day and painfully apparent in hindsight. A publish that quietly misplaced 40% of its site visitors over 1 / 4 doesn’t journey any alarm, however it hurts you over the lengthy run.
Right here’s the right way to automate it. As soon as 1 / 4, get Agent A to match each weblog publish’s natural site visitors towards the prior interval, flag those trending down, and write a refresh suggestion for every: what’s seemingly inflicting the drop and what to replace.
Floor the evaluation in actual numbers by connecting Ahrefs Net Analytics and pulling in Search Console knowledge by means of Ahrefs’ GSC Insights report.
This course of is out there in Agent A as a Talent or an App. The Declining Content material Detection ability does the decay evaluation, whereas the Weblog Freshness app offers you a standing dashboard of refresh candidates with the diagnostics connected.
Ask Agent A to arrange a schedule for this so it runs robotically with out you having to immediate it.
When you’d prefer to construct your individual, right here’s a starter immediate:
Starter immediate
Construct me a quarterly content-decay job. Pull natural site visitors for each weblog publish for the present quarter and the prior one. Flag posts the place site visitors dropped meaningfully. For every, write a refresh suggestion (seemingly trigger + what to replace) and append it as a brand new row in my Notion editorial database. Kind by site visitors misplaced so the largest decays are on the high.
You too can use Agent A to construct out a pipeline the place you’ll be able to paste an URL from the above device and replace them. That is precisely what our Director of Content material Advertising and marketing, Ryan Regulation did.
The pipeline fetches the article and run 5 diagnostics:
- Scope steerage — Set whether or not this can be a mild refresh or a full rewrite.
- Claims audit — the LLM flags each statistic, research reference, and dated assertion within the publish, grades every for staleness, and the place wanted suggests a alternative URL.
- Ahrefs mentions — cross-checks the article towards options launched since publication and suggests the place to say the brand new ones.
- Matter gaps — re-runs the SERP towards present top-ranking pages and surfaces subjects they cowl that mine doesn’t.
- Authoritative pages — finds new linkable sources printed because the article was printed.
The ultimate stage is a side-by-side diff between the present article and the proposed updates, with settle for/reject per change.
Right here’s a starter immediate to construct the replace pipeline:
Starter immediate
Construct me a blog-post replace pipeline. Enter: a broadcast URL. Fetch the article. Run 5 diagnostic phases: (1) Steering — I set scope (mild refresh vs. full rewrite); (2) Claims Audit — LLM extracts each stat, research reference, and dated assertion and grades every for staleness with a urged alternative; (3) Ahrefs Mentions — cross-check towards Ahrefs options launched since publication and counsel the place to drop new ones; (4) Matter Gaps — re-run the SERP, floor subjects present top-ranking pages cowl that mine doesn’t; (5) Authoritative Pages — discover linkable sources printed since my article. Last stage: side-by-side diff between present article and proposed updates, with settle for/reject per change. Export the accepted model as markdown and WordPress shortcodes.
Each AI writing device sells you an identical trick: a completed draft in two minutes. What they don’t point out is the remainder of the week that you simply then should spend fixing it. The final mile is the place the time went all alongside, and a quicker first draft doesn’t contact it.
Right here’s why: “Write me an article” is one immediate doing the work of 9 jobs without delay, and it does all 9 badly. Actual search engine optimisation content material is a sequence: key phrase analysis, SERP evaluation, hole evaluation, define, draft, inner hyperlinks, citations, photos, formatting. Collapse that right into a single ask and you may really feel each skipped step within the output.
So Ryan ran the chain as an alternative of the immediate. He constructed the Weblog Pipeline on Agent A: 11 phases, a goal key phrase in a single finish, a publish-ready draft out the different.
The distinction is that you simply see, and may edit, each stage because it goes. Agent A does the meeting; you retain the judgment. You cease salvaging AI drafts and begin approving them.
Starter immediate
Construct me an assisted long-form article pipeline. Atomic enter is a goal key phrase. Levels run sequentially as background jobs the UI polls: (1) key phrase analysis through Ahrefs, (2) competitor SERP fetch, (3) AI Content material Helper subject snapshot, (4) bulleted define with mandated subject protection, (5) data-mention placement, (6) full draft, (7) polish, (8) WordPress shortcode formatting + .docx export. Every stage reveals its output, has an “edit” textarea, and a “refine with suggestions” chat that re-runs the stage with my notes. Fashion information comes from a per-author voice profile.
Each group needs the month-to-month efficiency report. No one needs to make it.
So Ryan used Agent A to put in writing it. Now, on the first of every month, Agent A builds a weblog efficiency report and pulls the numbers collectively from Ahrefs Net Analytics and Ahrefs’ GSC Insights report in order that they keep constant.
Crucially, it isn’t traffic-only. The report covers the complete search engine optimisation image: clicks and impressions, key phrase rankings and place adjustments, referring domains and backlinks gained or misplaced, plus KPI tiles, month-over-month tendencies, and high movers. That’s what makes it an search engine optimisation report reasonably than a site visitors report.
Sooner or later, I can foresee Ryan piping this straight into our #weblog Slack channel through the Slack integration.
You are able to do precisely the identical as Ryan by putting in the Month-to-month Web site Efficiency Report app, out there in Agent A. Level it at your property, set the schedule, and decide your Slack channel.
When you’d prefer to construct your individual, right here’s a starter immediate:
Starter immediate
Construct me a month-to-month weblog report job. On the first of every month, pull GSC + Ahrefs Net Analytics for the prior month, compute KPI tiles (clicks, impressions, high gainers, high losers) with month-over-month deltas, render a brief abstract, and publish it to my #advertising Slack channel. Embrace a hyperlink to the complete report.
Each month, your rivals earn hyperlinks you’d like to have. Most of these are hyperlink alternatives sitting in plain sight.
Right here’s how we will automate them in Agent A.
As soon as a month, get Agent A to take a look at your foremost rivals, discover their finest new referring domains (sorted by the referring area’s personal site visitors and Area Ranking) and log them as a prospecting record.
These are potential prospects. The purpose is to see which publications and pages are linking out in our house proper now, so outreach goes to locations that demonstrably hyperlink to websites like ours. A high-DR website that simply linked a competitor is a far hotter goal than a chilly record.
You are able to do the identical through the use of the Hyperlink Intersect Prospecting ability to search out domains linking to rivals however to not you.
When you’re particularly on the lookout for damaged hyperlink constructing alternatives or linkbait alternatives, now we have these abilities out there too.
Run these abilities, then get Agent A to set it up as an automation.
Or when you’d prefer to construct your individual, right here’s a starter immediate:
Starter immediate
Construct me a month-to-month competitor backlink job. For my record of competitor domains, fetch new referring domains gained within the final month through Website Explorer, kind by the referring area’s personal site visitors then Area Ranking, drop low-quality and known-spam sources, and append the highest outcomes to an Airtable base with columns for supply URL, referring area, DR, site visitors, and which competitor it hyperlinks to.
AI assistants are quietly turning into a discovery channel, and what they are saying about you isn’t all the time what you’d say about your self.
Optimizing for the way you seem in AI solutions, whether or not you name it AEO or GEO, is more and more the search engine optimisation group’s job, as a result of it’s the identical work: understanding what surfaces win visibility, and incomes your home in them.
Right here’s the right way to automate it.
Ask Agent A to run a weekly evaluation of your model mentions throughout the key AI assistants (or simply the one which issues most to you). Agent A checks how the model is being described, then logs any recurring destructive themes (fallacious pricing, an outdated function declare, a competitor framed as higher).
When you use Notion or Airtable, you’ll be able to pipe them in there, too, so you’ll be able to observe patterns over time.
You’ll be able to simply do that in Agent A through the use of the AI Model Sentiment ability to run the immediate panel and observe framing over time. Then, pair it with the AI Point out Hole Evaluation ability to see the queries the place rivals get cited and also you don’t.
Ask Agent A to set it up as an automation for you and also you’re all set. (Or Agent A will even ask you if you need the automation.)
When you’d prefer to construct your individual, right here’s a starter immediate:
Starter immediate
Construct me a weekly brand-mention monitor for ChatGPT. Every week, test how my model is described in AI solutions through Model Radar, extract recurring themes, and log destructive or inaccurate ones (fallacious pricing, outdated claims, unfavorable competitor framing) into an Airtable base with the theme, an instance quote, and the date. Solely report AI share-of-voice relative to my named rivals.
Reddit is consuming the SERPs. There’s a very good likelihood a Reddit thread now outranks your rigorously optimized web page for a few of your key phrases.
Right here’s how one can automate this in Agent A.
As soon as every week, get Agent A to scan your high key phrases for newly-ranking Reddit discussions value partaking with. It checks the SERP for every precedence key phrase, isolates the Reddit outcomes that just lately broke into the highest positions, and passes you a shortlist.
The worth is timing. Leaping right into a thread that’s actively rating, whereas it’s contemporary and the query is stay, is value excess of discovering it six months later.
Starter immediate
Construct me a weekly Reddit-opportunity scanner. For my record of high key phrases, pull the present SERP through Key phrases Explorer / SERP overview, isolate reddit.com outcomes rating within the high 10, and flag ones that newly entered the highest positions this week. Give me a digest: key phrase, thread title and URL, present place, and the query being requested, sorted by key phrase quantity.
Everybody agrees inner linking ought to occur on each publish. Nearly no one does it on each publish.
It’s the flossing of search engine optimisation.
To repair this, Ryan constructed the Inner Linker. Give it a brand new article (a broadcast URL, or simply pasted draft markdown if it isn’t stay but) and it tells you which of them of your present posts ought to hyperlink to it.
The matching isn’t keyword-guesswork. It vector embeds your article with Gemini and cosine-compares it towards each different publish on the sitemap, then re-scores the highest matches by site visitors, so a related publish pulling actual natural site visitors ranks above an equally related one which no one visits. A hyperlink solely helps if it sits on a web page with authority to cross.
It additionally reads every candidate’s markdown and quietly drops any publish that already hyperlinks to you, so the record is simply locations you haven’t coated but.
It even finds the only paragraph that most closely fits your new article, and has Agent A write a pure 2-6 phrase anchor and rewrite that one sentence to incorporate it. You copy the rewritten sentence straight into the outdated publish. No deciding the place the hyperlink goes, no wording it your self.
Right here’s a starter immediate to recreate this for your self:
Starter immediate
Construct me an internal-linking device. Enter: both a broadcast weblog URL or pasted draft markdown for unpublished items. Embed the enter article with Gemini and cosine-compare towards my pre-cached weblog publish vectors. Rescore high candidates with authority weighting: 0.7 × similarity + 0.3 × log(org_traffic). Auto-exclude any host already linking to me (parse every candidate’s markdown physique). For every high host, determine the only paragraph most semantically aligned with the enter article, and that’s the place the hyperlink goes. Have Claude draft a pure 2-6 phrase anchor and rewrite a sentence within the host paragraph to incorporate it. Cache passage vectors per host so repeat lookups are on the spot.
Dropping a hyperlink from a DR 80 website issues. Dropping it silently, and discovering out 1 / 4 later when rankings slip, is the issue.
Right here’s how one can automate this in Agent A.
Each Monday, get Agent A to summarize your just lately misplaced backlinks from high-authority domains. When you have a devoted outreach particular person (or whoever’s in cost), you’ll be able to even get Agent A to ping them robotically in Slack (through the combination.)
It pulls misplaced backlinks from Website Explorer, filters to referring domains with a Area Ranking of fifty or greater, and contains the misplaced URL plus the web page it used to stay on, so reclamation outreach can begin the identical day.
Right here’s a starter immediate to construct this your self:
Starter immediate
Construct me a Monday lost-backlink alert. Every Monday, pull backlinks misplaced within the final week through Website Explorer, filter to referring domains with DR 50+, drop nofollow-only and known-spam losses. For every: the linking web page, the misplaced goal URL, the referring area’s DR, and the anchor textual content that was used.
Glen’s analysis discovered that “better of” lists are 43.8% of the pages ChatGPT cites. So, getting added to a listing like “finest search engine optimisation instruments” is a large win. However these lists change continually, and your spot in them issues, so a quiet demotion can value as a lot as removing.
Right here’s how we will use Agent A to repair this.
Everytime you uncover a brand new listicle point out, ask Agent A so as to add it to Firehose for monitoring. Then ask Agent A to re-check every tracked web page for adjustments each two weeks: did our entry get eliminated, did we slide down the rating, did a competitor get added above us?
It makes use of Web page Examine to diff the web page towards its final snapshot, so that you solely hear about it when one thing really modified
Starter immediate
Construct me a listicle-mention monitor. After I add a URL the place my model is talked about, register it in Firehose for monitoring. Each fortnight, re-fetch every tracked web page, diff it towards the final snapshot, and flag solely significant adjustments: my point out eliminated, my rank place modified, or a competitor added. Ship me a digest of simply the modified pages with a earlier than/after of the related part.
Last ideas
When you’re an Ahrefs buyer, you’ll be able to attempt Agent A without cost for one month.
Launch any of the talked about abilities, set up the apps, or paste the starter prompts right into a contemporary workspace and your individual Agent A will begin constructing, or seize the completed instruments from the applying library.
